Is India Keeping Up?

All releases

Every tracked LLM release since 1 January 2026, newest first.

  • All months
  • October 2026
  • September 2026
  • August 2026
  • July 2026
  • June 2026
  • May 2026
  • April 2026
  • March 2026
  • February 2026
  • January 2026

October 2026

1 model
LabModelsReleaseDateSource
Reflection AIUS1BeamText-only sparse MoE (501B total, 23B active) for coding, reasoning and agentic work, pretrained on 23.8T tokens. Announced Oct 5 with Apache 2.0 weights promised later in October; early access via waitlist.Official announcementReflection announcement post
  • Reflection AIUS

    Beam

September 2026

80 models
LabModelsReleaseDateSource
Ant Group (inclusionAI)CN1Ling-3.1-flashAPI-available via Novita on Vercel AI Gateway at announcement; open weights promised but not yet released. Gateway metadata lists a September 29 provider date.Official announcementVercel AI Gateway listing
Google / Google DeepMindUS1Gemini 4 ArgonFrontier model announced Sep 30; initially limited to trusted cyber defenders via Google's Fairwind Program, with API (paid tier) and Google AI Ultra access promised later.Official announcementGoogle DeepMind announcement post
OpenAIUS1GPT-6.1 SolOfficial announcementOpenAI announcement post
AnthropicUS1Claude Sonnet 5.5Official announcementAnthropic announcement post
VIDRAFT / FINAL-BenchKR1Darwin-180B-RSIQwen3.8-180B reasoning derivative; benchmark claims are publisher reports, not independently audited here.Model card and weightsPublisher repository history
VIDRAFT / FINAL-BenchKR1Darwin-27B-RSIQwen3.8-27B reasoning derivative with publisher-described recursive self-improvement.Model card and weightsPublisher repository history
Xiaomi MiMoCN2MiMo-V2.6-Pro-MOPD, MiMo-V2.6-Flash-MOPDOfficial announcementOfficial model cardOfficial model card
MeituanCN1LongCat-2.5-PreviewOfficial announcement
PerceptronUS1Perceptron Mk1.5Official announcement
FreedomIntelligence / CUHK ShenzhenCN1HuatuoGPT-3-27BDate combines dated developer release documentation with first complete-weight availability. Historical public/private visibility was not independently archived.Official announcementOfficial model card
Fireworks AIUS1Ember-1Official announcementOfficial announcement
AnthropicUS1Claude Opus 5.5Official announcement
AppleUS1LensVLM-9BDate combines first complete weights with contemporaneous developer documentation; historical repository visibility was not independently archived.Official announcementOfficial model card
OpenAIUS2GPT-6 Sol, GPT-6 LunaOfficial announcement
UpstageKR1Solar Mini 4 (solar-mini4-260922)Public proprietary Solar API language model; native output is text.Official model release historyOfficial model release history
Xiaomi MiMoCN1MiMo-V2.6-Distill-Qwen-9BOfficial announcementOfficial model card
Xiaomi MiMoCN2MiMo-V2.6-Pro-RL, MiMo-V2.6-Flash-RLOfficial announcement
xAIUS1Grok 4.7Official announcement
StepFunCN1Step 5 PreviewPublic launch September 20; earlier September 18 tracker/benchmark date is not evidence of public access.Official announcementOfficial announcement
AI SingaporeSG2Nemotron-SEA-LION-v4.8-30B-A3B, Nemotron-SEA-LION-v4.8-120B-A12BSoutheast Asian language adaptation of NVIDIA Nemotron 3 Nano/Super. Base and post-trained checkpoints are distinct; precision variants are counted once. Model card and weightsModel card and weightsPublisherTechnical paper
AI SingaporeSG2Nemotron-SEA-LION-v4.8-30B-A3B-Base, Nemotron-SEA-LION-v4.8-120B-A12B-BaseSoutheast Asian language adaptation of NVIDIA Nemotron 3 Nano/Super. Base and post-trained checkpoints are distinct; precision variants are counted once. Model card and weightsModel card and weightsPublisherTechnical paper
Trillion LabsKR1Trida2.0-4BQwen3.5-4B adapted to block-diffusion text generation. A later repository rename does not add a release.Model card and weightsPublisher repository history
Alibaba (Qwen)CN1Qwen3.8-Omni-FlashOfficial announcement
FreedomIntelligence / CUHK ShenzhenCN1HuatuoGPT-3-9BDate combines dated developer release documentation with first complete-weight availability. Historical public/private visibility was not independently archived.Official announcementOfficial model card
Liquid AIUS2LFM2-1.2B-Longevity, LFM2-2.6B-LongevitySeptember 17 dated developer announcement; repositories were created earlier, and an earlier public-availability date has not been established.Official announcement
PrismMLUS1Ternary-Bonsai-2-27BCounted as a separately trained native low-bit model; equivalent precision/packaging conversions are excluded.Official announcement
Ant Group (inclusionAI)CN1LLaDA-UIDate uses dated developer release documentation together with complete public-weight history. Repository history does not independently prove the exact private-to-public transition.Official announcementOfficial model card
PoSTMEDIA AI LabKR1Xin-V1Korean instruction model post-trained from Qwen3.8-27B.Model card and weightsPublisher repository history
Shanghai AI LaboratoryCN1Atria Dawn PreviewOfficial announcementOfficial announcementOfficial model card
Shanghai AI LaboratoryCN1Intern-S2-397BOfficial announcementOfficial model card
PaletteKR1songgot-lTiny Korean-first causal language model for textual function calls, trained from scratch. Counted despite small size; later training updates in the same named repository are not separate releases.Model card and weightsPublisherPublisher repository history
PaletteKR1songgot-x-0.8bQwen3.5-0.8B post-trained for Korean textual tool calling. Distinct from the from-scratch Songgot family.Model card and weightsPublisher repository history
NVIDIAUS1NV-Reason-CTDate uses dated developer release documentation together with complete public-weight history. Repository history does not independently prove the exact private-to-public transition.Official announcementOfficial model card
PoSTMEDIA AI LabKR1Lux-V2Second-generation Korean instruction model derived from Gemma 4 26B A4B.Model card and weightsPublisher repository history
PoSTMEDIA AI LabKR1Lux-V2-ProSecond-generation Korean instruction model derived from Gemma 4 31B.Model card and weightsPublisher repository history
Moonshot AICN1Kimi K2.8 PreviewOfficial announcement
CognitionUS1SWE-2Official announcement
DeepSeekCN1DeepSeek-V4.1-FlashOfficial announcement
NVIDIAUS1NemotronLabs-AI-for-Media-Sports-TennisOfficial model card
PaletteKR1songgotTiny Korean-first causal language model for textual function calls, trained from scratch. Counted despite small size; later training updates in the same named repository are not separate releases.Model card and weightsPublisherPublisher repository history
PaletteKR1songgot-12lTiny Korean-first causal language model for textual function calls, trained from scratch. Counted despite small size; later training updates in the same named repository are not separate releases.Model card and weightsPublisherPublisher repository history
PaletteKR1songgot-mTiny Korean-first causal language model for textual function calls, trained from scratch. Counted despite small size; later training updates in the same named repository are not separate releases.Model card and weightsPublisherPublisher repository history
PoSTMEDIA AI LabKR1Rosetta-7B-ThinkReasoning checkpoint of Rosetta 7B.Model card and weightsPublisher repository history
Trillion LabsKR1Gravity-30B-A5B-BaseExpanded 30B/5B-active base language model using block expansion and additional training from Gravity 16B.Model card and weightsPublisher repository history
Trillion LabsKR1Gravity-30B-A5B-PreviewInstruction/reasoning preview of the 30B Gravity family.Model card and weightsPublisher repository history
InceptionUS1Mercury 2.5Official announcementOfficial announcement
Lunit ConsortiumKR1L1-30B-A5BMedical reasoning and tool-assisted clinical conversation derived from the Gravity 30B family.Model card and weightsPublisher repository history
Nex AGICN3Nex-N2.5-mini, Nex-N2.5-Pro, Nex-N2.5-MaxDate uses the dated developer release revision together with access links and weight history; original public visibility is not independently archived. Pro first appears via hosted access, with weights two days later.Official announcementOfficial model cardOfficial model cardOfficial model card
PoSTMEDIA AI LabKR1Rosetta-7B-InstructInstruction checkpoint of Rosetta 7B.Model card and weightsPublisher repository history
OpenBMB / ModelBestCN1MiniCPM5-2BOfficial announcementOfficial model card
OpenBMB / ModelBestCN2MiniCPM5-2B-Base, MiniCPM5-2B-SFTUses the publisher-stated September 7 family launch for the Base/SFT artifacts present in the release table; weights and preparatory documentation existed September 5.Official announcementOfficial model cardOfficial model card
PoSTMEDIA AI LabKR1Rosetta-7B-Base7B general Korean language base model. Training and ancestry claims remain publisher-reported.Model card and weightsPublisher repository history
Ant Group (inclusionAI)CN1LLaDA2.2-miniDate uses complete public-weight history; no separate dated announcement was found.Official model card
PoSTMEDIA AI LabKR1Lux-V1-Pro-Heritage-KoCultural-heritage domain language model with Korean conversational text output.Model card and weightsPublisher repository history
Ant Group (inclusionAI)CN1Ling-3.0-flash-VLOfficial release notes
Ant Group (inclusionAI)CN1Ling-3.0-flash-SanteHealthcare-focused post-train of Ling-3.0-flash, available via API (OpenRouter, Vercel AI Gateway, Novita); no public weights found.Official announcementOpenRouter listing
OpenAIUS1GPT-6 AstraAnnounced September 3 with restricted access; public paid-plan/API availability is documented September 4 UTC.Official announcementOfficial release notesOpenAI public-access announcement
Ant Group (inclusionAI)CN1Ling-3.0-flash-FinOfficial announcementOfficial model card
NVIDIAUS2Nemotron-3-Labs-Ultra-Math-RL, Nemotron-3-Labs-Ultra-Math-SFTOfficial model cardOfficial model card
Alibaba (Qwen)CN1Qwen-Drive-1.0-4BSeptember 2 is the earliest verified dated public usage documentation after the manuscript described code as forthcoming; the exact first-public weight transition is not recorded.Official announcementOfficial model card
Alibaba (Qwen)CN1Qwen3.8-Max-0902Official announcement
Google / Google DeepMindUS1Gemini 3.8 FlashOfficial announcement
Meta AIUS1Muse Spark 1.3Official announcement
NVIDIAUS4SDLLM-AR-1.7B-Base, SDLLM-MDLM-1.7B-Base, SDLLM-Duo-1.7B-Base, SDLLM-EsoLM-1.7B-BaseOfficial model cardOfficial model cardOfficial model cardOfficial model card
AnthropicUS1Claude Fable 5.1Official announcementDated API release notes
iFLYTEKCN2Spark-X2.5-4B, Spark-X2.5-1.7BOfficial announcement
  • Ant Group (inclusionAI)CN

    Ling-3.1-flash

  • Google / Google DeepMindUS

    Gemini 4 Argon

  • OpenAIUS

    GPT-6.1 Sol

  • AnthropicUS

    Claude Sonnet 5.5

  • VIDRAFT / FINAL-BenchKR

    Darwin-180B-RSI

  • VIDRAFT / FINAL-BenchKR

    Darwin-27B-RSI

  • Xiaomi MiMoCN

    MiMo-V2.6-Pro-MOPD · MiMo-V2.6-Flash-MOPD

  • MeituanCN

    LongCat-2.5-Preview

  • PerceptronUS

    Perceptron Mk1.5

  • FreedomIntelligence / CUHK ShenzhenCN

    HuatuoGPT-3-27B

  • Fireworks AIUS

    Ember-1

  • AnthropicUS

    Claude Opus 5.5

  • AppleUS

    LensVLM-9B

  • OpenAIUS

    GPT-6 Sol · GPT-6 Luna

  • UpstageKR

    Solar Mini 4 (solar-mini4-260922)

  • Xiaomi MiMoCN

    MiMo-V2.6-Distill-Qwen-9B

  • Xiaomi MiMoCN

    MiMo-V2.6-Pro-RL · MiMo-V2.6-Flash-RL

  • xAIUS

    Grok 4.7

  • StepFunCN

    Step 5 Preview

  • AI SingaporeSG

    Nemotron-SEA-LION-v4.8-30B-A3B · Nemotron-SEA-LION-v4.8-120B-A12B

  • AI SingaporeSG

    Nemotron-SEA-LION-v4.8-30B-A3B-Base · Nemotron-SEA-LION-v4.8-120B-A12B-Base

  • Trillion LabsKR

    Trida2.0-4B

  • Alibaba (Qwen)CN

    Qwen3.8-Omni-Flash

  • FreedomIntelligence / CUHK ShenzhenCN

    HuatuoGPT-3-9B

  • Liquid AIUS

    LFM2-1.2B-Longevity · LFM2-2.6B-Longevity

  • PrismMLUS

    Ternary-Bonsai-2-27B

  • Ant Group (inclusionAI)CN

    LLaDA-UI

  • PoSTMEDIA AI LabKR

    Xin-V1

  • Shanghai AI LaboratoryCN

    Atria Dawn Preview

  • Shanghai AI LaboratoryCN

    Intern-S2-397B

  • PaletteKR

    songgot-l

  • PaletteKR

    songgot-x-0.8b

  • NVIDIAUS

    NV-Reason-CT

  • PoSTMEDIA AI LabKR

    Lux-V2

  • PoSTMEDIA AI LabKR

    Lux-V2-Pro

  • Moonshot AICN

    Kimi K2.8 Preview

  • CognitionUS

    SWE-2

  • DeepSeekCN

    DeepSeek-V4.1-Flash

  • NVIDIAUS

    NemotronLabs-AI-for-Media-Sports-Tennis

  • PaletteKR

    songgot

  • PaletteKR

    songgot-12l

  • PaletteKR

    songgot-m

  • PoSTMEDIA AI LabKR

    Rosetta-7B-Think

  • Trillion LabsKR

    Gravity-30B-A5B-Base

  • Trillion LabsKR

    Gravity-30B-A5B-Preview

  • InceptionUS

    Mercury 2.5

  • Lunit ConsortiumKR

    L1-30B-A5B

  • Nex AGICN

    Nex-N2.5-mini · Nex-N2.5-Pro · Nex-N2.5-Max

  • PoSTMEDIA AI LabKR

    Rosetta-7B-Instruct

  • OpenBMB / ModelBestCN

    MiniCPM5-2B

  • OpenBMB / ModelBestCN

    MiniCPM5-2B-Base · MiniCPM5-2B-SFT

  • PoSTMEDIA AI LabKR

    Rosetta-7B-Base

  • Ant Group (inclusionAI)CN

    LLaDA2.2-mini

  • PoSTMEDIA AI LabKR

    Lux-V1-Pro-Heritage-Ko

  • Ant Group (inclusionAI)CN

    Ling-3.0-flash-VL

  • Ant Group (inclusionAI)CN

    Ling-3.0-flash-Sante

  • OpenAIUS

    GPT-6 Astra

  • Ant Group (inclusionAI)CN

    Ling-3.0-flash-Fin

  • NVIDIAUS

    Nemotron-3-Labs-Ultra-Math-RL · Nemotron-3-Labs-Ultra-Math-SFT

  • Alibaba (Qwen)CN

    Qwen-Drive-1.0-4B

  • Alibaba (Qwen)CN

    Qwen3.8-Max-0902

  • Google / Google DeepMindUS

    Gemini 3.8 Flash

  • Meta AIUS

    Muse Spark 1.3

  • NVIDIAUS

    SDLLM-AR-1.7B-Base · SDLLM-MDLM-1.7B-Base · SDLLM-Duo-1.7B-Base · SDLLM-EsoLM-1.7B-Base

  • AnthropicUS

    Claude Fable 5.1

  • iFLYTEKCN

    Spark-X2.5-4B · Spark-X2.5-1.7B

August 2026

59 models
LabModelsReleaseDateSource
Tencent HunyuanCN3ContextPilot-8B, ContextPilot-14B, ContextPilot-E4BDate uses dated developer release documentation together with complete public-weight history. Repository history does not independently prove the exact private-to-public transition.Official announcementOfficial model cardOfficial model cardOfficial model card
Tencent HunyuanCN1Hy4 PreviewOfficial announcement
Alibaba (Qwen)CN1Qwen3.8-Flash-Next (Qwen3.8-Flash)Official announcementOfficial model card
Ant Group (inclusionAI)CN1UI-Venus-2-9BDate uses dated developer release documentation together with complete public-weight history. Repository history does not independently prove the exact private-to-public transition.Official announcementOfficial model card
Z.ai / Zhipu AICN1GLM-5.3-FlashOfficial release notesOfficial model cardOfficial announcement
ApodexUS1Apodex-1.1-miniMini was announced August 24; August 25 is the first complete downloadable checkpoint in the inspected repository history.Official announcementOfficial model card
IBMUS3Granite-4.2-3b, Granite-4.2-8b, Granite-4.2-30bOfficial model cardOfficial model cardOfficial model card
ApodexUS1Apodex 1.1Official announcement
PaletteKR1palette-k-14bQwen2.5-14B instruction model adapted for Korean enterprise document drafting and business correspondence.Model card and weightsPublisher repository history
DeepSeekCN1DeepSeek-V4-Flash-Vision-ExpOfficial announcement
Ant Group (inclusionAI)CN2Ling-3.0-tiny-base, Ling-3.0-flash-baseOfficial announcementOfficial model cardOfficial model card
OpenBMB / ModelBestCN1MathForm-8BOfficial announcement
Tencent HunyuanCN3UI-Mate-9B, UI-Mate-27B, UI-Mate-democua-27BDate uses dated developer release documentation together with complete public-weight history. Repository history does not independently prove the exact private-to-public transition.Official announcementOfficial model cardOfficial model cardOfficial model card
Alibaba (Qwen)CN1Qwen3.8-27BOfficial announcementOfficial model card
NVIDIAUS5Nemotron-Labs-Teacher-Competition-Coding, Nemotron-Labs-Teacher-Chat, Nemotron-Labs-Teacher-Instruction-Following, Nemotron-Labs-Teacher-General-Reasoning, Nemotron-Labs-Teacher-STEMOfficial model cardOfficial model cardOfficial model cardOfficial model cardOfficial model card
Xiaohongshu / rednoteCN1dots3-note-previewOfficial announcementOfficial model cardOfficial announcement
Z.ai / Zhipu AICN1GLM-5.3Official announcementOfficial release notes
DeepSeekCN1DeepSeek-V4-Pro-0813Official announcement
Google / Google DeepMindUS1Gemini 3.7 FlashOfficial announcement
Liquid AIUS1LFM2.5-VL-3BOfficial announcement
MicrosoftUS1MAI-Thinking-1Public preview August 12; June 2 technical-report date is not used as launch.Official announcement
xAIUS1Grok 4.6Official announcement
MicrosoftUS1MAI-Code-1.1-FlashOfficial announcement
NVIDIAUS2Nemotron-3.5-Lightning-30B-A3B, Nemotron-3.5-Lightning-30B-A3B-BaseOfficial model cardOfficial model card
Gnani.aiIN1Gnani Evon v3.3 30B-A3B30B hybrid Mamba-Transformer MoE (~3.5B active), continued pretraining for 11 Indian languages, open weights under Apache 2.0.Hugging FaceGnani.ai
Meta AIUS1Muse GlimmerOfficial announcement
Motif TechnologiesKR1Motif-3-BaseUnaligned base checkpoint of Motif 3, separate from beta and final instruction models.Model card and weightsPublisher repository history
Motif TechnologiesKR1Motif-3Final Motif 3 general reasoning/agentic checkpoint; distinct from the beta.Model card and weightsPublisherPublisher repository historyEarlier publisher weight distributionPublisher repository history
Allen Institute for AI (Ai2)US12intent-aware-lfqa-llama3-8b-baseline, intent-aware-lfqa-llama3-8b-intent-explicit, intent-aware-lfqa-llama3-8b-intent-implicit, intent-aware-lfqa-llama3-8b-multiview, intent-aware-lfqa-qwen3-4b-baseline, intent-aware-lfqa-qwen3-4b-intent-explicit, intent-aware-lfqa-qwen3-4b-intent-implicit, intent-aware-lfqa-qwen3-4b-multiview, intent-aware-lfqa-qwen3-8b-baseline, intent-aware-lfqa-qwen3-8b-intent-explicit, intent-aware-lfqa-qwen3-8b-intent-implicit, intent-aware-lfqa-qwen3-8b-multiviewOfficial announcement
Ant Group (inclusionAI)CN1Ling-3.0-tinyOfficial announcementOfficial release notes
UpstageKR1Solar Pro 4 (solar-pro4-260806)Public proprietary Solar API language model; native output is text.Official model release historyOfficial model release history
Meta AIUS1Muse Spark 1.2Official announcement
Shanghai AI LaboratoryCN1Intern-S2-MobiusFirst dated developer release documentation, corroborated by complete weights. The earlier August 4 upload alone does not establish public visibility.Official announcementOfficial model card
Liquid AIUS2LFM2.5-2.6B, LFM2.5-2.6B-BaseOfficial announcement
MeituanCN1LongCat-Flash-Lite-SparseDate is the first verified dated developer statement explicitly releasing this trained variant, after complete weights; exact earlier public visibility is not independently logged.Official announcementOfficial model card
  • Tencent HunyuanCN

    ContextPilot-8B · ContextPilot-14B · ContextPilot-E4B

  • Tencent HunyuanCN

    Hy4 Preview

  • Alibaba (Qwen)CN

    Qwen3.8-Flash-Next (Qwen3.8-Flash)

  • Ant Group (inclusionAI)CN

    UI-Venus-2-9B

  • Z.ai / Zhipu AICN

    GLM-5.3-Flash

  • ApodexUS

    Apodex-1.1-mini

  • IBMUS

    Granite-4.2-3b · Granite-4.2-8b · Granite-4.2-30b

  • ApodexUS

    Apodex 1.1

  • PaletteKR

    palette-k-14b

  • DeepSeekCN

    DeepSeek-V4-Flash-Vision-Exp

  • Ant Group (inclusionAI)CN

    Ling-3.0-tiny-base · Ling-3.0-flash-base

  • OpenBMB / ModelBestCN

    MathForm-8B

  • Tencent HunyuanCN

    UI-Mate-9B · UI-Mate-27B · UI-Mate-democua-27B

  • Alibaba (Qwen)CN

    Qwen3.8-27B

  • NVIDIAUS

    Nemotron-Labs-Teacher-Competition-Coding · Nemotron-Labs-Teacher-Chat · Nemotron-Labs-Teacher-Instruction-Following · Nemotron-Labs-Teacher-General-Reasoning · Nemotron-Labs-Teacher-STEM

  • Xiaohongshu / rednoteCN

    dots3-note-preview

  • Z.ai / Zhipu AICN

    GLM-5.3

  • DeepSeekCN

    DeepSeek-V4-Pro-0813

  • Google / Google DeepMindUS

    Gemini 3.7 Flash

  • Liquid AIUS

    LFM2.5-VL-3B

  • MicrosoftUS

    MAI-Thinking-1

  • xAIUS

    Grok 4.6

  • MicrosoftUS

    MAI-Code-1.1-Flash

  • NVIDIAUS

    Nemotron-3.5-Lightning-30B-A3B · Nemotron-3.5-Lightning-30B-A3B-Base

  • Gnani.aiIN

    Gnani Evon v3.3 30B-A3B

  • Meta AIUS

    Muse Glimmer

  • Motif TechnologiesKR

    Motif-3-Base

  • Motif TechnologiesKR

    Motif-3

  • Allen Institute for AI (Ai2)US

    intent-aware-lfqa-llama3-8b-baseline · intent-aware-lfqa-llama3-8b-intent-explicit · intent-aware-lfqa-llama3-8b-intent-implicit · intent-aware-lfqa-llama3-8b-multiview · intent-aware-lfqa-qwen3-4b-baseline · intent-aware-lfqa-qwen3-4b-intent-explicit · intent-aware-lfqa-qwen3-4b-intent-implicit · intent-aware-lfqa-qwen3-4b-multiview · intent-aware-lfqa-qwen3-8b-baseline · intent-aware-lfqa-qwen3-8b-intent-explicit · intent-aware-lfqa-qwen3-8b-intent-implicit · intent-aware-lfqa-qwen3-8b-multiview

  • Ant Group (inclusionAI)CN

    Ling-3.0-tiny

  • UpstageKR

    Solar Pro 4 (solar-pro4-260806)

  • Meta AIUS

    Muse Spark 1.2

  • Shanghai AI LaboratoryCN

    Intern-S2-Mobius

  • Liquid AIUS

    LFM2.5-2.6B · LFM2.5-2.6B-Base

  • MeituanCN

    LongCat-Flash-Lite-Sparse

July 2026

59 models
LabModelsReleaseDateSource
DeepSeekCN1DeepSeek-V4-Flash-0731Official announcement
HuaweiCN1openPangu-2.0-ProOfficial announcementOfficial announcement
LG AI ResearchKR1K-EXAONE-2.0-750B-A37B750B/37B-active MoE language model in LG’s sovereign-model program.Model card and weightsPublisherPublisher
Google / Google DeepMindUS1Gemini Robotics ER 2Official announcement
Thinking Machines LabUS1Inkling-SmallOfficial announcement
SK TelecomKR1A.X-K2688B MoE language model with enhanced reasoning, tool use and long-context capabilities.Model card and weightsPublisherPublisher
KakaoKR2kanana-2-1.3b-base, kanana-2-3b-baseSmall Korean language models for general text use; distinct base and instruction checkpoints. Model card and weightsModel card and weightsPublisher
KakaoKR2kanana-2-1.3b-instruct, kanana-2-3b-instructSmall Korean language models for general text use; distinct base and instruction checkpoints. Subject to the publisher’s model license.Model card and weightsModel card and weightsPublisher
Liquid AIUS1LFM2.5-Encoder-350M-DiffusionOfficial announcementOfficial model card
MicrosoftUS1Mage-VLOfficial announcementOfficial model card
Trillion LabsKR1sim-scholar-qwen3-4bS3 research agent: Qwen3-4B trained with reinforcement learning for tool-assisted scientific literature search and text answers.Model card and weightsPublisherPublisher repository history
AnthropicUS1Claude Opus 5Official announcement
Nanbeige LLM Lab / BOSS ZhipinCN1Nanbeige4.2-3BFirst dated developer release statement verified is July 24; earlier HF creation alone does not establish public availability.Official announcementOfficial model cardOfficial organization attribution
Ant Group (inclusionAI)CN1Ling-3.0-flashOfficial announcementOfficial model card
Kuaishou (Kwaipilot)CN1KAT-Coder-V2.5-DevJuly 23 is supported by the complete-weight upload and accompanying developer release card, corroborated at month precision by the project lead. Historical private-to-public visibility is not separately logged.Official model cardOfficial model cardOfficial announcement
NVIDIAUS1Ising-Calibration-1.5-31BOfficial model card
Ant Group (inclusionAI)CN1LLaDA2.2-flashDate uses dated developer release documentation together with complete public-weight history. Repository history does not independently prove the exact private-to-public transition.Official announcementOfficial model card
JD.comCN1JoyAI-VL-InteractionOfficial announcement
UpstageKR1Solar-Open2-250B250B/15B-active agentic MoE expanded from Solar Open using selective weight transfer and additional training.Model card and weightsPublisher
Alibaba (Qwen)CN1Qwen3.7-FlashOfficial announcement
Google / Google DeepMindUS2Gemini 3.6 Flash, Gemini 3.5 Flash-LiteOfficial announcement
PoolsideUS1Laguna S 2.1Official announcement
VIDRAFT / FINAL-BenchKR1AETHER-7B-7Attn-baseExperimental causal language model with mixed attention mechanisms. Publisher describes training from scratch.Model card and weightsPublisher repository history
VIDRAFT / FINAL-BenchKR1Aether-6B-11Attn-baseExperimental causal language model with mixed attention mechanisms. Publisher describes training from scratch.Model card and weightsPublisher repository history
Alibaba (Qwen)CN1Qwen3.8-Max (Qwen3.8-2.4T-A95B)Preview-to-GA counted once absent explicit evidence of a new trained checkpoint.Official announcementOfficial announcementOfficial model card
VIDRAFT / FINAL-BenchKR1Aether-7B-5Attn-itExperimental causal language model with mixed attention mechanisms. Instruction-tuned derivative of Aether 7B.Model card and weightsPublisher repository history
Shanghai AI LaboratoryCN1Intern-S2-Preview-397BDate is the first dated developer release documentation corroborated by complete model weights; historical repository visibility was not independently archived.Official announcementOfficial model card
VIDRAFT / FINAL-BenchKR1Aether-7B-5AttnExperimental causal language model with mixed attention mechanisms. Publisher describes training from scratch.Model card and weightsPublisher repository history
Moonshot AICN1Kimi K3Official announcementOfficial announcement
xAIUS1Grok 4.5Official announcement
Tencent HunyuanCN1Hy-Embodied-VLM-1.0Official model card
Thinking Machines LabUS1InklingOfficial announcement
Motif TechnologiesKR1Motif-3-Beta300B-class MoE beta made usable through Chat Motif before the later weight release.Model card and weightsPublisher
OpenMOSS / Fudan UniversityCN3MOSS-VL-Base-0708, MOSS-VL-Instruct-0708, MOSS-VL-RealtimeOfficial announcementOfficial model cardOfficial model cardOfficial model card
PrismMLUS2Bonsai-27B, Ternary-Bonsai-27BCounted as a separately trained native low-bit model; equivalent precision/packaging conversions are excluded.Official announcement
Tencent HunyuanCN1HiLS-Attention-7BDate uses dated developer release documentation together with complete public-weight history. Repository history does not independently prove the exact private-to-public transition.Official announcementOfficial model card
VIDRAFT / FINAL-BenchKR1Ourbox-35B-JGOSGeneral text/reasoning checkpoint created by combining existing language-model weights. This is a derivative release, not evidence of a newly pretrained foundation or independently verified benchmark leadership.Model card and weightsPublisher repository history
Meta AIUS1Muse Spark 1.1Official announcement
OpenAIUS3GPT-5.6 Sol, GPT-5.6 Terra, GPT-5.6 LunaOfficial announcement
CognitionUS1SWE-1.7Official announcement
IBMUS2Granite-swash-2b, Granite-swash-3b-a600mOfficial model cardOfficial model card
Kuaishou (Kwaipilot)CN1KAT-Coder-Air-V2.5July 7 is the first dated developer API listing located; first-day access is not independently archived.Official announcementOfficial announcement
Kuaishou (Kwaipilot)CN1KAT-Coder-V2.5 (Pro)July 6 is the earliest dated primary availability statement located; third-party hosting dates are later.Official announcementDeveloper API catalog and dated pricing
NVIDIAUS1Nemotron-Labs-3-Puzzle-75B-A9BOfficial model card
Tencent HunyuanCN1Hy3Official announcement
VIDRAFT / FINAL-BenchKR1JGOS-31B-Citizen-RGeneral text/reasoning checkpoint created by combining existing language-model weights. This is a derivative release, not evidence of a newly pretrained foundation or independently verified benchmark leadership.Model card and weightsPublisher repository history
VIDRAFT / FINAL-BenchKR1Darwin-4B-ChimeraPublic Chimera research checkpoint with hybrid transplanted architecture.Model card and weightsPublisher repository history
MicrosoftUS2HARC-Qwen2.5-7B-Instruct, HARC-Llama-3.1-8B-InstructDate uses dated developer release documentation together with complete public-weight history. Repository history does not independently prove the exact private-to-public transition.Official announcementOfficial model cardOfficial model card
PoolsideUS1Laguna XS 2.1Official announcement
  • DeepSeekCN

    DeepSeek-V4-Flash-0731

  • HuaweiCN

    openPangu-2.0-Pro

  • LG AI ResearchKR

    K-EXAONE-2.0-750B-A37B

  • Google / Google DeepMindUS

    Gemini Robotics ER 2

  • Thinking Machines LabUS

    Inkling-Small

  • SK TelecomKR

    A.X-K2

  • KakaoKR

    kanana-2-1.3b-base · kanana-2-3b-base

  • KakaoKR

    kanana-2-1.3b-instruct · kanana-2-3b-instruct

  • Liquid AIUS

    LFM2.5-Encoder-350M-Diffusion

  • MicrosoftUS

    Mage-VL

  • Trillion LabsKR

    sim-scholar-qwen3-4b

  • AnthropicUS

    Claude Opus 5

  • Nanbeige LLM Lab / BOSS ZhipinCN

    Nanbeige4.2-3B

  • Ant Group (inclusionAI)CN

    Ling-3.0-flash

  • Kuaishou (Kwaipilot)CN

    KAT-Coder-V2.5-Dev

  • NVIDIAUS

    Ising-Calibration-1.5-31B

  • Ant Group (inclusionAI)CN

    LLaDA2.2-flash

  • JD.comCN

    JoyAI-VL-Interaction

  • UpstageKR

    Solar-Open2-250B

  • Alibaba (Qwen)CN

    Qwen3.7-Flash

  • Google / Google DeepMindUS

    Gemini 3.6 Flash · Gemini 3.5 Flash-Lite

  • PoolsideUS

    Laguna S 2.1

  • VIDRAFT / FINAL-BenchKR

    AETHER-7B-7Attn-base

  • VIDRAFT / FINAL-BenchKR

    Aether-6B-11Attn-base

  • Alibaba (Qwen)CN

    Qwen3.8-Max (Qwen3.8-2.4T-A95B)

  • VIDRAFT / FINAL-BenchKR

    Aether-7B-5Attn-it

  • Shanghai AI LaboratoryCN

    Intern-S2-Preview-397B

  • VIDRAFT / FINAL-BenchKR

    Aether-7B-5Attn

  • Moonshot AICN

    Kimi K3

  • xAIUS

    Grok 4.5

  • Tencent HunyuanCN

    Hy-Embodied-VLM-1.0

  • Thinking Machines LabUS

    Inkling

  • Motif TechnologiesKR

    Motif-3-Beta

  • OpenMOSS / Fudan UniversityCN

    MOSS-VL-Base-0708 · MOSS-VL-Instruct-0708 · MOSS-VL-Realtime

  • PrismMLUS

    Bonsai-27B · Ternary-Bonsai-27B

  • Tencent HunyuanCN

    HiLS-Attention-7B

  • VIDRAFT / FINAL-BenchKR

    Ourbox-35B-JGOS

  • Meta AIUS

    Muse Spark 1.1

  • OpenAIUS

    GPT-5.6 Sol · GPT-5.6 Terra · GPT-5.6 Luna

  • CognitionUS

    SWE-1.7

  • IBMUS

    Granite-swash-2b · Granite-swash-3b-a600m

  • Kuaishou (Kwaipilot)CN

    KAT-Coder-Air-V2.5

  • Kuaishou (Kwaipilot)CN

    KAT-Coder-V2.5 (Pro)

  • NVIDIAUS

    Nemotron-Labs-3-Puzzle-75B-A9B

  • Tencent HunyuanCN

    Hy3

  • VIDRAFT / FINAL-BenchKR

    JGOS-31B-Citizen-R

  • VIDRAFT / FINAL-BenchKR

    Darwin-4B-Chimera

  • MicrosoftUS

    HARC-Qwen2.5-7B-Instruct · HARC-Llama-3.1-8B-Instruct

  • PoolsideUS

    Laguna XS 2.1

June 2026

53 models
LabModelsReleaseDateSource
AnthropicUS1Claude Sonnet 5Official announcement
HuaweiCN1openPangu-2.0-FlashOfficial announcementOfficial announcement
MeituanCN1LongCat-2.0Official page date June 30; discovery dataset uses June 29. The dated publisher page takes precedence.Official announcementOfficial model card
NVIDIAUS1Nemotron-Labs-Audio-Visual-Flamingo-InstructOfficial model card
Trillion LabsKR1Gravity-bio-16B-A3BBiology-domain language model continued from Gravity-16B-A3B; general domain text generation.Model card and weightsPublisher repository history
Liquid AIUS2LFM2.5-230M, LFM2.5-230M-BaseOfficial announcement
NVIDIAUS1Nemotron-Labs-TwoTower-30B-A3B-BaseUses the first verified dated developer statement explicitly releasing the model; complete artifact history predates the statement and does not establish earlier public visibility.Official announcementOfficial model card
Alibaba (Qwen)CN1Qwen-AgentWorld-35B-A3BOfficial announcement
ByteDance SeedCN2Seed2.1 Pro, Seed2.1 TurboOfficial announcement
VIDRAFT / FINAL-BenchKR1Darwin-36B-KRKorean instruction tuning of the Qwen3.6 35B MoE.Model card and weightsPublisher repository history
Allen Institute for AI (Ai2)US1Tmax-27BUses the dated developer paper explicitly releasing models after complete 27B weights became available; exact earlier visibility is not independently logged.Official announcementOfficial model card
JD.comCN1JoyAI-VL-Interaction-PreviewOfficial announcement
VIDRAFT / FINAL-BenchKR1Darwin-Chimera-4B-Gen1Qwen3-4B backbone with attention healing training and frozen feed-forward layers.Model card and weightsPublisher repository history
VIDRAFT / FINAL-BenchKR1Darwin-V9-Chimera-4BCross-architecture feed-forward transplant followed by healing training.Model card and weightsPublisher repository history
VIDRAFT / FINAL-BenchKR1Darwin-V9-Chimera-4B-SFTInstruction-tuned Chimera Gen2 checkpoint.Model card and weightsPublisher repository history
Allen Institute for AI (Ai2)US4Tmax-2B, Tmax-4B, Tmax-8B, Tmax-9BDate uses dated developer release documentation together with complete public-weight history. Repository history does not independently prove the exact private-to-public transition.Official announcementOfficial model cardOfficial model cardOfficial model cardOfficial model card
Z.ai / Zhipu AICN1GLM-5.2Official release notesOfficial model cardOfficial announcement
VIDRAFT / FINAL-BenchKR1Darwin-397B-JGOSGeneral text/reasoning checkpoint created by combining existing language-model weights. This is a derivative release, not evidence of a newly pretrained foundation or independently verified benchmark leadership.Model card and weightsPublisher repository history
VIDRAFT / FINAL-BenchKR1JGOS-31B-Citizen-v2Korean public-service conversation model fine-tuned with citizen-service instruction data.Model card and weightsPublisher repository history
Moonshot AICN1Kimi K2.7 CodeOfficial announcement
Google / Google DeepMindUS1DiffusionGemma 26B-A4B ITOfficial announcementOfficial model card
Alibaba (Qwen)CN1Qwen3.7-Max-2026-06-08Official announcement
AnthropicUS1Claude Fable 5Official announcement
NVIDIAUS1CUDA-AutocompleteOfficial model card
ApodexUS1Apodex 1.0Official announcement
ApodexUS4Apodex-1.0-mini, Apodex-1.0-0.8B-SFT, Apodex-1.0-2B-SFT, Apodex-1.0-4B-SFTOfficial announcement
AppleUS4AFM 3 Core, AFM 3 Core Advanced, AFM 3 Cloud, AFM 3 Cloud ProFour Apple Foundation Models with text output, available through the developer beta on supported devices and Private Cloud Compute. Uses the publicly accessible developer-beta launch of the AFM 3-powered features. Apple does not expose a separate rollout timestamp or selectable endpoint for every routed variant.Official announcementOfficial announcementOfficial announcementOfficial announcement
UnisoundCN1U2Official announcementOfficial announcement
Nex AGICN2Nex-N2-Pro, Nex-N2-miniDate follows the June 4 developer release documentation and complete artifacts; original repository visibility is not independently archived.Official announcementOfficial announcementOfficial model cardOfficial model card
NVIDIAUS2Nemotron-3-Ultra-550B-A55B, Nemotron-3-Ultra-550B-A55B-BaseOfficial model cardOfficial model card
AI SingaporeSG1Llama3.2-1B-DuDiSmall conversational checkpoint trained with DuDi cross-lingual distillation; a research release, not a SEA-LION rebrand.Model card and weightsTechnical paperPublisher repository history
AI SingaporeSG1Qwen2.5-0.5B-DuDiSmall conversational checkpoint trained with DuDi cross-lingual distillation; a research release, not a SEA-LION rebrand.Model card and weightsTechnical paperPublisher repository history
AI SingaporeSG1Qwen2.5-1.5B-DuDiSmall conversational checkpoint trained with DuDi cross-lingual distillation; a research release, not a SEA-LION rebrand.Model card and weightsTechnical paperPublisher repository history
AI SingaporeSG1Qwen3-0.6B-Base-DuDiSmall conversational checkpoint trained with DuDi cross-lingual distillation; a research release, not a SEA-LION rebrand.Model card and weightsTechnical paperPublisher repository history
Google / Google DeepMindUS2Gemma 4 12B Base, Gemma 4 12B ITOfficial announcement
MicrosoftUS1MAI-Code-1-FlashOfficial announcement
Alibaba (Qwen)CN1Qwen3.7-PlusOfficial announcement
MiniMaxCN1MiniMax M3Official announcement
NVIDIAUS14D-RGPT-8BPublisher states June 1 release; the currently discovered Hugging Face repository was created June 2, so exact first-access day is qualified.Official model card
  • AnthropicUS

    Claude Sonnet 5

  • HuaweiCN

    openPangu-2.0-Flash

  • MeituanCN

    LongCat-2.0

  • NVIDIAUS

    Nemotron-Labs-Audio-Visual-Flamingo-Instruct

  • Trillion LabsKR

    Gravity-bio-16B-A3B

  • Liquid AIUS

    LFM2.5-230M · LFM2.5-230M-Base

  • NVIDIAUS

    Nemotron-Labs-TwoTower-30B-A3B-Base

  • Alibaba (Qwen)CN

    Qwen-AgentWorld-35B-A3B

  • ByteDance SeedCN

    Seed2.1 Pro · Seed2.1 Turbo

  • VIDRAFT / FINAL-BenchKR

    Darwin-36B-KR

  • Allen Institute for AI (Ai2)US

    Tmax-27B

  • JD.comCN

    JoyAI-VL-Interaction-Preview

  • VIDRAFT / FINAL-BenchKR

    Darwin-Chimera-4B-Gen1

  • VIDRAFT / FINAL-BenchKR

    Darwin-V9-Chimera-4B

  • VIDRAFT / FINAL-BenchKR

    Darwin-V9-Chimera-4B-SFT

  • Allen Institute for AI (Ai2)US

    Tmax-2B · Tmax-4B · Tmax-8B · Tmax-9B

  • Z.ai / Zhipu AICN

    GLM-5.2

  • VIDRAFT / FINAL-BenchKR

    Darwin-397B-JGOS

  • VIDRAFT / FINAL-BenchKR

    JGOS-31B-Citizen-v2

  • Moonshot AICN

    Kimi K2.7 Code

  • Google / Google DeepMindUS

    DiffusionGemma 26B-A4B IT

  • Alibaba (Qwen)CN

    Qwen3.7-Max-2026-06-08

  • AnthropicUS

    Claude Fable 5

  • NVIDIAUS

    CUDA-Autocomplete

  • ApodexUS

    Apodex 1.0

  • ApodexUS

    Apodex-1.0-mini · Apodex-1.0-0.8B-SFT · Apodex-1.0-2B-SFT · Apodex-1.0-4B-SFT

  • AppleUS

    AFM 3 Core · AFM 3 Core Advanced · AFM 3 Cloud · AFM 3 Cloud Pro

  • UnisoundCN

    U2

  • Nex AGICN

    Nex-N2-Pro · Nex-N2-mini

  • NVIDIAUS

    Nemotron-3-Ultra-550B-A55B · Nemotron-3-Ultra-550B-A55B-Base

  • AI SingaporeSG

    Llama3.2-1B-DuDi

  • AI SingaporeSG

    Qwen2.5-0.5B-DuDi

  • AI SingaporeSG

    Qwen2.5-1.5B-DuDi

  • AI SingaporeSG

    Qwen3-0.6B-Base-DuDi

  • Google / Google DeepMindUS

    Gemma 4 12B Base · Gemma 4 12B IT

  • MicrosoftUS

    MAI-Code-1-Flash

  • Alibaba (Qwen)CN

    Qwen3.7-Plus

  • MiniMaxCN

    MiniMax M3

  • NVIDIAUS

    4D-RGPT-8B

May 2026

60 models
LabModelsReleaseDateSource
StepFunCN1Step-3.7-FlashOfficial announcementOfficial model card
AnthropicUS1Claude Opus 4.8Official announcement
Liquid AIUS2LFM2.5-8B-A1B, LFM2.5-8B-A1B-BaseOfficial announcement
UpstageKR1TinySolar-111m-4kSmall causal language-model research checkpoint with weights, tokenizer and 4K-context configuration; training lineage is not documented.Model card and weightsPublisher repository history
UpstageKR1TinySolar-187m-4kSmall causal language-model research checkpoint with weights, tokenizer and 4K-context configuration; training lineage is not documented.Model card and weightsPublisher repository history
Essential AIUS1Rnj-1.5-InstructOfficial announcementOfficial announcementOfficial model card
OpenBMB / ModelBestCN2MiniCPM5-1B-Base, MiniCPM5-1B-SFTDate uses dated developer release documentation together with complete public-weight history. Repository history does not independently prove the exact private-to-public transition.Official announcementOfficial model cardOfficial model card
VIDRAFT / FINAL-BenchKR1Darwin-218B-ExpertGeneral text/reasoning checkpoint created by combining existing language-model weights. This is a derivative release, not evidence of a newly pretrained foundation or independently verified benchmark leadership.Model card and weightsPublisher repository history
VIDRAFT / FINAL-BenchKR1Darwin-218B-DelphiChemistry/scientific language adaptation, originally named Darwin-218B-Expert-Chem.Model card and weightsPublisher repository history
VIDRAFT / FINAL-BenchKR1Darwin-218B-Expert-BioBiology-domain instruction/reasoning adaptation.Model card and weightsPublisher repository history
VIDRAFT / FINAL-BenchKR1Darwin-218B-krGeneral text/reasoning checkpoint created by combining existing language-model weights. This is a derivative release, not evidence of a newly pretrained foundation or independently verified benchmark leadership.Model card and weightsPublisher repository history
MicrosoftUS3Fara1.5-4B, Fara1.5-9B, Fara1.5-27BOfficial announcement
AI SingaporeSG2Qwen-SEA-LION-v4.5-27B-IT, Gemma-SEA-LION-v4.5-E2B-ITRegional multilingual instruction models for agentic use. Gemma accepts visual/audio inputs but its released decoder generates text.Model card and weightsModel card and weightsPublisher
Alibaba (Qwen)CN1Qwen3.7-MaxOfficial announcement
VIDRAFT / FINAL-BenchKR1Darwin-28B-CoderCoding adaptation of Darwin 28B.Model card and weightsPublisher repository history
Google / Google DeepMindUS1Gemini 3.5 FlashOfficial announcement
NVIDIAUS7Nemotron-Labs-Diffusion-3B, Nemotron-Labs-Diffusion-3B-Base, Nemotron-Labs-Diffusion-8B, Nemotron-Labs-Diffusion-8B-Base, Nemotron-Labs-Diffusion-14B, Nemotron-Labs-Diffusion-14B-Base, Nemotron-Labs-Diffusion-VLM-8BOfficial announcementOfficial announcementOfficial model cardOfficial model cardOfficial model cardOfficial model cardOfficial model cardOfficial model cardOfficial model card
OpenBMB / ModelBestCN1MiniCPM5-1BDeveloper release history dates May 19; Hugging Face repository creation is May 21.Official announcementOfficial model card
Cursor / AnysphereUS1Composer 2.5Official announcement
PoSTMEDIA AI LabKR1Lux-V1Korean instruction adaptation of Gemma 4 26B A4B.Model card and weightsPublisher repository history
PoSTMEDIA AI LabKR1Lux-V1-ProKorean instruction adaptation of Gemma 4 31B.Model card and weightsPublisher repository history
VIDRAFT / FINAL-BenchKR1Darwin-28B-REASONReasoning-transplant-distillation language checkpoint; external test-time reasoning is not counted as another model.Model card and weightsPublisher repository history
VIDRAFT / FINAL-BenchKR1Darwin-Mistral-G1General text/reasoning checkpoint created by combining existing language-model weights. This is a derivative release, not evidence of a newly pretrained foundation or independently verified benchmark leadership.Model card and weightsTechnical paperPublisher repository history
ByteDance SeedCN1Cola-DLMMay 15 is the earliest verified dated developer release documentation with complete weights; the earlier paper date alone is not used.Official announcementOfficial model card
Shanghai AI LaboratoryCN1Intern-S2-PreviewDate is the first dated developer release documentation corroborated by complete model weights; historical repository visibility was not independently archived.Official announcementOfficial model card
VIDRAFT / FINAL-BenchKR1Darwin-28B-KOREAGeneral text/reasoning checkpoint created by combining existing language-model weights. This is a derivative release, not evidence of a newly pretrained foundation or independently verified benchmark leadership.Model card and weightsPublisher repository history
VIDRAFT / FINAL-BenchKR1Darwin-9B-KOREAGeneral text/reasoning checkpoint created by combining existing language-model weights. This is a derivative release, not evidence of a newly pretrained foundation or independently verified benchmark leadership.Model card and weightsPublisher repository history
MicrosoftUS1MagenticBrainOfficial model card
xAIUS1Grok Build 0.1Model identity linked retrospectively by the May 29 API announcement; first public paid application access was May 14.Official release notesOfficial announcement
DeCLaRe LabSG1delta-mem_qwen3_4b-instructPublic delta-Mem memory adapter and inference code for Qwen3-4B-Instruct-2507, enabling long-context language generation. Base model is required; this counts the adapted model once.Model card and weightsPublisherPublisher repository history
OpenBMB / ModelBestCN2MiniCPM-V 4.6, MiniCPM-V 4.6 ThinkingOfficial announcementOfficial model cardOfficial model card
VIDRAFT / FINAL-BenchKR1Darwin-9B-NEG-x-Negentropy-V8General text/reasoning checkpoint created by combining existing language-model weights. This is a derivative release, not evidence of a newly pretrained foundation or independently verified benchmark leadership.Model card and weightsPublisher repository history
BaiduCN1ERNIE 5.1Official announcementOfficial announcement
Ant Group (inclusionAI)CN1Ring-2.6-1TOfficial announcementOfficial announcement
NVIDIAUS1Nemotron-Labs-3-Elastic-30B-A3BOfficial model card
Allen Institute for AI (Ai2)US7Emo_1b14b_1T, StdMoE_1b14b_1T, Emo_1b14b_130B, StdMoE_1b14b_130B, Dense_1b_130B, StdMoE_1b4b_130B, StdMoE_1b14b_1T_EmoAnnealedUses the first verified dated developer table containing the full suite; most weights were uploaded April 29 and the final annealed variant May 5. Exact earlier repository visibility is not independently logged.Official announcementOfficial model cardOfficial model cardOfficial model cardOfficial model cardOfficial model cardOfficial model cardOfficial model cardOfficial model card
IBMUS3Granite-switch-4.1-3b-preview, Granite-switch-4.1-8b-preview, Granite-switch-4.1-30b-previewOfficial model cardOfficial model cardOfficial model card
OpenAIUS1GPT-5.5 InstantOfficial announcement
OpenMOSS / Fudan UniversityCN2MOSS-Music-8B-Instruct, MOSS-Music-8B-ThinkingOfficial model cardOfficial model card
  • StepFunCN

    Step-3.7-Flash

  • AnthropicUS

    Claude Opus 4.8

  • Liquid AIUS

    LFM2.5-8B-A1B · LFM2.5-8B-A1B-Base

  • UpstageKR

    TinySolar-111m-4k

  • UpstageKR

    TinySolar-187m-4k

  • Essential AIUS

    Rnj-1.5-Instruct

  • OpenBMB / ModelBestCN

    MiniCPM5-1B-Base · MiniCPM5-1B-SFT

  • VIDRAFT / FINAL-BenchKR

    Darwin-218B-Expert

  • VIDRAFT / FINAL-BenchKR

    Darwin-218B-Delphi

  • VIDRAFT / FINAL-BenchKR

    Darwin-218B-Expert-Bio

  • VIDRAFT / FINAL-BenchKR

    Darwin-218B-kr

  • MicrosoftUS

    Fara1.5-4B · Fara1.5-9B · Fara1.5-27B

  • AI SingaporeSG

    Qwen-SEA-LION-v4.5-27B-IT · Gemma-SEA-LION-v4.5-E2B-IT

  • Alibaba (Qwen)CN

    Qwen3.7-Max

  • VIDRAFT / FINAL-BenchKR

    Darwin-28B-Coder

  • Google / Google DeepMindUS

    Gemini 3.5 Flash

  • NVIDIAUS

    Nemotron-Labs-Diffusion-3B · Nemotron-Labs-Diffusion-3B-Base · Nemotron-Labs-Diffusion-8B · Nemotron-Labs-Diffusion-8B-Base · Nemotron-Labs-Diffusion-14B · Nemotron-Labs-Diffusion-14B-Base · Nemotron-Labs-Diffusion-VLM-8B

  • OpenBMB / ModelBestCN

    MiniCPM5-1B

  • Cursor / AnysphereUS

    Composer 2.5

  • PoSTMEDIA AI LabKR

    Lux-V1

  • PoSTMEDIA AI LabKR

    Lux-V1-Pro

  • VIDRAFT / FINAL-BenchKR

    Darwin-28B-REASON

  • VIDRAFT / FINAL-BenchKR

    Darwin-Mistral-G1

  • ByteDance SeedCN

    Cola-DLM

  • Shanghai AI LaboratoryCN

    Intern-S2-Preview

  • VIDRAFT / FINAL-BenchKR

    Darwin-28B-KOREA

  • VIDRAFT / FINAL-BenchKR

    Darwin-9B-KOREA

  • MicrosoftUS

    MagenticBrain

  • xAIUS

    Grok Build 0.1

  • DeCLaRe LabSG

    delta-mem_qwen3_4b-instruct

  • OpenBMB / ModelBestCN

    MiniCPM-V 4.6 · MiniCPM-V 4.6 Thinking

  • VIDRAFT / FINAL-BenchKR

    Darwin-9B-NEG-x-Negentropy-V8

  • BaiduCN

    ERNIE 5.1

  • Ant Group (inclusionAI)CN

    Ring-2.6-1T

  • NVIDIAUS

    Nemotron-Labs-3-Elastic-30B-A3B

  • Allen Institute for AI (Ai2)US

    Emo_1b14b_1T · StdMoE_1b14b_1T · Emo_1b14b_130B · StdMoE_1b14b_130B · Dense_1b_130B · StdMoE_1b4b_130B · StdMoE_1b14b_1T_EmoAnnealed

  • IBMUS

    Granite-switch-4.1-3b-preview · Granite-switch-4.1-8b-preview · Granite-switch-4.1-30b-preview

  • OpenAIUS

    GPT-5.5 Instant

  • OpenMOSS / Fudan UniversityCN

    MOSS-Music-8B-Instruct · MOSS-Music-8B-Thinking

April 2026

132 models
LabModelsReleaseDateSource
Ant Group (inclusionAI)CN1Ling-2.6-1TUses documented April 29 platform launch; an earlier April 23 third-party hosting lead remains unresolved.Official release notes
IBMUS6Granite-4.1-3b, Granite-4.1-8b, Granite-4.1-30b, Granite-4.1-3b-base, Granite-4.1-8b-base, Granite-4.1-30b-baseOfficial model cardOfficial model cardOfficial model cardOfficial model cardOfficial model cardOfficial model card
NVIDIAUS1Cosmos-Reason2-32BOfficial model card
NVIDIAUS1Nemotron-3-Nano-Omni-30B-A3B-ReasoningOfficial model card
PoolsideUS2Laguna XS.2, Laguna M.1Official announcement
GiniGen AIKR1Rogue-28B-MIXKorean language derivative merging Darwin and Quetta model ancestry. Training-free merging is documented separately from the four-value provenance vocabulary.Model card and weightsPublisher repository history
Xiaomi MiMoCN2MiMo-V2.5-Base, MiMo-V2.5-Pro-BaseOfficial announcementOfficial model cardOfficial model card
VIDRAFT / FINAL-BenchKR1Darwin-28B-KRGeneral text/reasoning checkpoint created by combining existing language-model weights. This is a derivative release, not evidence of a newly pretrained foundation or independently verified benchmark leadership.Model card and weightsPublisher repository history
VIDRAFT / FINAL-BenchKR1Darwin-28B-KR-LegalKorean legal reasoning and conversation adaptation of Darwin 28B.Model card and weightsPublisher repository history
AppleUS1CADD-Base-7BDate combines first complete weights with contemporaneous developer documentation; historical repository visibility was not independently archived.Official announcementOfficial model card
DeepSeekCN2DeepSeek-V4-Flash Preview, DeepSeek-V4-Pro PreviewOfficial announcement
Tencent HunyuanCN1HY-Embodied-0.5-XOfficial announcement
VIDRAFT / FINAL-BenchKR1Darwin-28B-OpusGeneral text/reasoning checkpoint created by combining existing language-model weights. This is a derivative release, not evidence of a newly pretrained foundation or independently verified benchmark leadership.Model card and weightsPublisher repository history
Alibaba (Qwen)CN1Qwen3.5-Plus-2026-04-20Official announcement
Allen Institute for AI (Ai2)US26OlmPool H_pre_LQK_32kv_8k_11k_SWA, OlmPool G_pre_8kv_8k_14k_SWA, OlmPool H_post_LQK_8kv_4k_11k_SWA, OlmPool G_post_LQK_8kv_4k_14k_SWA, OlmPool B_post_LQK_32kv_4k_11k, OlmPool K_post_HQK_8kv_12k, OlmPool H_post_LQK_32kv_8k_11k_SWA, OlmPool A_post_HQK_8kv_8k_13k_SWA_fp8, OlmPool H_post_LQK_32kv_8k_11k_SWA_fp8, OlmPool F_pre_8kv_8k_14k_SWA, OlmPool E_post_LQK_32kv_8k_11k_SWA_fp8, OlmPool H_post_HQK_32kv_8k_11k_SWA, OlmPool C_post_LQK_8kv_8k_13k_SWA, OlmPool D_post_LQK_8kv_8k_13k_SWA_fp8, OlmPool J_pre_16kv_8k_14k, OlmPool G_pre_8kv_8k_14k, OlmPool I_pre_32kv_8k_12k, OlmPool G_post_LQK_8kv_8k_14k, OlmPool G_pre_LQK_8kv_8k_14k, OlmPool H_pre_32kv_8k_11k_SWA, OlmPool G_pre_4kv_8k_14k, OlmPool G_pre_8kv_4k_14k, OlmPool H_post_LQK_32kv_8k_11k, OlmPool H_post_LQK_32kv_4k_11k, OlmPool B_post_LQK_32kv_4k_11k_SWA, OlmPool D_post_LQK_8kv_8k_13k_SWAUse the April 23 publisher release of original OLMo-core checkpoints. Later Hugging Face conversion/upload dates do not create additional releases.Official announcementOfficial announcementOfficial model card
OpenAIUS2GPT-5.5, GPT-5.5 ProOfficial announcement
Tencent HunyuanCN2Hy3 Preview, Hy3 Preview BaseOfficial model cardOfficial model card
Xiaomi MiMoCN1MiMo-V2.5-ProOfficial announcement
Alibaba (Qwen)CN1Qwen3.6-27BOfficial announcementOfficial model card
Ant Group (inclusionAI)CN2DR-Venus-4B-SFT, DR-Venus-4B-RLOfficial announcement
Ant Group (inclusionAI)CN1Ling-2.6-flashOfficial release notesOfficial announcement
VIDRAFT / FINAL-BenchKR1Darwin-36B-OpusGeneral text/reasoning checkpoint created by combining existing language-model weights. This is a derivative release, not evidence of a newly pretrained foundation or independently verified benchmark leadership.Model card and weightsPublisher repository history
Xiaomi MiMoCN1MiMo-V2.5April 22 model page versus April 23 API log; both are documented.Official announcementOfficial announcement
ServiceNowUS2SuperApriel-15B-Base, SuperApriel-15B-InstructUses the April 21 dated developer release statement; repositories were created earlier, on April 7, without independently archived public-visibility dates.Official announcementOfficial model cardOfficial model card
Trillion LabsKR1NemoSlidesNemotron 3 Nano adaptation generating Slidev slide code/text; rendering occurs outside the language model.Model card and weightsPublisherPublisher repository history
Moonshot AICN1Kimi K2.6Official announcementOfficial announcementOfficial announcement
VIDRAFT / FINAL-BenchKR1Darwin-2B-OpusFinal v3 Korean reasoning checkpoint; lastbrain-v3 and the separately hosted adapter are aliases of the same adaptation.Model card and weightsPublisher repository history
VIDRAFT / FINAL-BenchKR1Qwen3.5-2B-Opus-Distill-v1Published instruction-distillation LoRA adapter for Qwen3.5-2B; base required.Model card and weightsPublisher repository history
VIDRAFT / FINAL-BenchKR1Qwen3.5-2B-Opus-SDPO-v1Published preference-optimization LoRA adapter for Qwen3.5-2B; base required.Model card and weightsPublisher repository history
VIDRAFT / FINAL-BenchKR1lastbrainInitial merged Qwen3.5-2B instruction-distillation experiment.Model card and weightsPublisher repository history
VIDRAFT / FINAL-BenchKR1lastbrain-v2Second instruction-tuning experiment with expanded math, code and Korean data.Model card and weightsPublisher repository history
xAIUS1Grok 4.3 BetaOfficial release notes
Alibaba (Qwen)CN1Qwen3.6-35B-A3B (Qwen3.6-Flash)Hosted Flash identity should be revisited if developer documents a distinct checkpoint.Official announcementOfficial announcement
AnthropicUS1Claude Opus 4.7Official announcement
PrismMLUS3Ternary-Bonsai-1.7B, Ternary-Bonsai-4B, Ternary-Bonsai-8BCounted as a separately trained native low-bit model; equivalent precision/packaging conversions are excluded.Official announcement
Tencent HunyuanCN1POINTS-SeekerOfficial announcementOfficial model card
VIDRAFT / FINAL-BenchKR1Darwin-27B-KR-V3Further Korean cultural/linguistic instruction tuning of the Darwin 27B family.Model card and weightsPublisher repository history
Alibaba (Qwen)CN1Qwen3.6-Max-PreviewOfficial announcement
Google / Google DeepMindUS1Gemini Robotics-ER 1.6Official announcement
NVIDIAUS1Ising-Calibration-1-35B-A3BOfficial model card
VIDRAFT / FINAL-BenchKR1Darwin-27B-KR-V2Korean SFT derivative of Darwin-27B-KR.Model card and weightsPublisher repository history
GiniGen AIKR1Rogue-27B-KRKorean instruction fine-tune derived from Darwin-27B-Opus.Model card and weightsPublisher repository history
NVIDIAUS3Audio-Flamingo-Next-Instruct, Audio-Flamingo-Next-Think, Audio-Flamingo-Next-CaptionerUses the April 13 dated developer release statement; the first day of repository public visibility is not independently archived.Official announcementOfficial model cardOfficial model cardOfficial model card
OpenMOSS / Fudan UniversityCN4MOSS-Audio-4B-Instruct, MOSS-Audio-4B-Thinking, MOSS-Audio-8B-Instruct, MOSS-Audio-8B-ThinkingOfficial announcementOfficial model cardOfficial model cardOfficial model cardOfficial model card
VIDRAFT / FINAL-BenchKR1Darwin-27B-KRGeneral text/reasoning checkpoint created by combining existing language-model weights. This is a derivative release, not evidence of a newly pretrained foundation or independently verified benchmark leadership.Model card and weightsTechnical paperPublisher repository history
VIDRAFT / FINAL-BenchKR1Darwin-27B-OpusGeneral text/reasoning checkpoint created by combining existing language-model weights. This is a derivative release, not evidence of a newly pretrained foundation or independently verified benchmark leadership.Model card and weightsPublisher research articleTechnical paper
VIDRAFT / FINAL-BenchKR1Darwin-4B-GenesisGeneral text/reasoning checkpoint created by combining existing language-model weights. This is a derivative release, not evidence of a newly pretrained foundation or independently verified benchmark leadership.Model card and weightsTechnical paperPublisher repository history
LG AI ResearchKR1EXAONE-4.5-33BVision-language model producing text from text and image input; extends EXAONE 4.0.Model card and weightsPublisherTechnical paper
Tencent HunyuanCN1HY-Embodied-0.5-MoT-2BPublisher explicitly dates open-weight release April 9; the currently inspected HF repository history starts April 14, so repository timestamps alone cannot recover the first availability transition.Official announcement
Trillion LabsKR1Gravity-16B-A3B-Base16B/3B-active causal MoE trained from scratch with the Lunit consortium.Model card and weightsPublisher repository history
Liquid AIUS1LFM2.5-VL-450MOfficial announcement
Lunit ConsortiumKR1L1-16B-A3BMedical conversation and clinical reasoning model derived from Gravity-16B-A3B-Base.Model card and weightsPublisher repository history
Meta AIUS1Muse SparkOfficial announcement
OpenMOSS / Fudan UniversityCN2MOSS-VL-Base-0408, MOSS-VL-Instruct-0408Official announcementOfficial model cardOfficial model card
VIDRAFT / FINAL-BenchKR1Darwin-4B-OpusGeneral text/reasoning checkpoint created by combining existing language-model weights. This is a derivative release, not evidence of a newly pretrained foundation or independently verified benchmark leadership.Model card and weightsTechnical paperPublisher repository history
AppleUS3SimpleSD-4B-instruct, SimpleSD-30B-instruct, SimpleSD-4B-thinkingOfficial announcementOfficial model cardOfficial model cardOfficial model card
CognitionUS1SWE-1.6Official announcement
NEWEN AIKR1QuettaLLMs-27B-Koreasoner-V3Qwen3.5-27B fine-tuned with Korean knowledge and logical reasoning instruction data.Model card and weightsPublisherPublisher repository history
PoSTMEDIA AI LabKR1Vayne-V3-Pro120B counterpart of Vayne V3 for Korean enterprise conversation and tool use.Model card and weightsPublisher repository history
Z.ai / Zhipu AICN1GLM-5.1Official release notesOfficial model cardOfficial announcement
VIDRAFT / FINAL-BenchKR1Darwin-31B-OpusGeneral text/reasoning checkpoint created by combining existing language-model weights. This is a derivative release, not evidence of a newly pretrained foundation or independently verified benchmark leadership.Model card and weightsTechnical paperPublisher repository history
VIDRAFT / FINAL-BenchKR1Darwin-9B-OpusGeneral text/reasoning checkpoint created by combining existing language-model weights. This is a derivative release, not evidence of a newly pretrained foundation or independently verified benchmark leadership.Model card and weightsTechnical paperPublisher repository history
AmazonUS9BMOJOF-primed-HQwen3-8B-Instruct, GDN-primed-HQwen3-32B-Instruct, GDN-primed-HQwen3-8B-Instruct, GDN-primed-HQwen3-8B-Reasoner, GKA-primed-HQwen3-32B-Instruct, GKA-primed-HQwen3-32B-Reasoner, GKA-primed-HQwen3-8B-Instruct, GKA-primed-HQwen3-8B-Reasoner, Mamba2-primed-HQwen3-8B-InstructConservative first dated developer documentation after complete weights: April 3. Complete artifacts were uploaded late April 2 UTC, but historical public visibility was not independently archived.Official announcementOfficial model cardOfficial model cardOfficial model cardOfficial model cardOfficial model cardOfficial model cardOfficial model cardOfficial model cardOfficial model card
JD.comCN2JoyAI-LLM-Flash, JoyAI-LLM-Flash-BaseUses the April 3 dated developer release statement. Repositories were created in February; that metadata alone does not establish an earlier public release.Official announcementOfficial model cardOfficial model card
Google / Google DeepMindUS8Gemma 4 E2B Base, Gemma 4 E2B IT, Gemma 4 E4B Base, Gemma 4 E4B IT, Gemma 4 26B-A4B Base, Gemma 4 26B-A4B IT, Gemma 4 31B Base, Gemma 4 31B ITOfficial announcementOfficial announcement
Alibaba (Qwen)CN1Qwen3.6-PlusOfficial announcement
Arcee AIUS1Trinity-Large-ThinkingOfficial announcement
Tencent HunyuanCN1Sequential-Hidden-Decoding-8B-n8-InstructOfficial announcement
Z.ai / Zhipu AICN1GLM-5V-TurboApril 1 UTC social launch; repository news uses April 2, consistent with China local date.Official announcementOfficial announcement
  • Ant Group (inclusionAI)CN

    Ling-2.6-1T

  • IBMUS

    Granite-4.1-3b · Granite-4.1-8b · Granite-4.1-30b · Granite-4.1-3b-base · Granite-4.1-8b-base · Granite-4.1-30b-base

  • NVIDIAUS

    Cosmos-Reason2-32B

  • NVIDIAUS

    Nemotron-3-Nano-Omni-30B-A3B-Reasoning

  • PoolsideUS

    Laguna XS.2 · Laguna M.1

  • GiniGen AIKR

    Rogue-28B-MIX

  • Xiaomi MiMoCN

    MiMo-V2.5-Base · MiMo-V2.5-Pro-Base

  • VIDRAFT / FINAL-BenchKR

    Darwin-28B-KR

  • VIDRAFT / FINAL-BenchKR

    Darwin-28B-KR-Legal

  • AppleUS

    CADD-Base-7B

  • DeepSeekCN

    DeepSeek-V4-Flash Preview · DeepSeek-V4-Pro Preview

  • Tencent HunyuanCN

    HY-Embodied-0.5-X

  • VIDRAFT / FINAL-BenchKR

    Darwin-28B-Opus

  • Alibaba (Qwen)CN

    Qwen3.5-Plus-2026-04-20

  • Allen Institute for AI (Ai2)US

    OlmPool H_pre_LQK_32kv_8k_11k_SWA · OlmPool G_pre_8kv_8k_14k_SWA · OlmPool H_post_LQK_8kv_4k_11k_SWA · OlmPool G_post_LQK_8kv_4k_14k_SWA · OlmPool B_post_LQK_32kv_4k_11k · OlmPool K_post_HQK_8kv_12k · OlmPool H_post_LQK_32kv_8k_11k_SWA · OlmPool A_post_HQK_8kv_8k_13k_SWA_fp8 · OlmPool H_post_LQK_32kv_8k_11k_SWA_fp8 · OlmPool F_pre_8kv_8k_14k_SWA · OlmPool E_post_LQK_32kv_8k_11k_SWA_fp8 · OlmPool H_post_HQK_32kv_8k_11k_SWA · OlmPool C_post_LQK_8kv_8k_13k_SWA · OlmPool D_post_LQK_8kv_8k_13k_SWA_fp8 · OlmPool J_pre_16kv_8k_14k · OlmPool G_pre_8kv_8k_14k · OlmPool I_pre_32kv_8k_12k · OlmPool G_post_LQK_8kv_8k_14k · OlmPool G_pre_LQK_8kv_8k_14k · OlmPool H_pre_32kv_8k_11k_SWA · OlmPool G_pre_4kv_8k_14k · OlmPool G_pre_8kv_4k_14k · OlmPool H_post_LQK_32kv_8k_11k · OlmPool H_post_LQK_32kv_4k_11k · OlmPool B_post_LQK_32kv_4k_11k_SWA · OlmPool D_post_LQK_8kv_8k_13k_SWA

  • OpenAIUS

    GPT-5.5 · GPT-5.5 Pro

  • Tencent HunyuanCN

    Hy3 Preview · Hy3 Preview Base

  • Xiaomi MiMoCN

    MiMo-V2.5-Pro

  • Alibaba (Qwen)CN

    Qwen3.6-27B

  • Ant Group (inclusionAI)CN

    DR-Venus-4B-SFT · DR-Venus-4B-RL

  • Ant Group (inclusionAI)CN

    Ling-2.6-flash

  • VIDRAFT / FINAL-BenchKR

    Darwin-36B-Opus

  • Xiaomi MiMoCN

    MiMo-V2.5

  • ServiceNowUS

    SuperApriel-15B-Base · SuperApriel-15B-Instruct

  • Trillion LabsKR

    NemoSlides

  • Moonshot AICN

    Kimi K2.6

  • VIDRAFT / FINAL-BenchKR

    Darwin-2B-Opus

  • VIDRAFT / FINAL-BenchKR

    Qwen3.5-2B-Opus-Distill-v1

  • VIDRAFT / FINAL-BenchKR

    Qwen3.5-2B-Opus-SDPO-v1

  • VIDRAFT / FINAL-BenchKR

    lastbrain

  • VIDRAFT / FINAL-BenchKR

    lastbrain-v2

  • xAIUS

    Grok 4.3 Beta

  • Alibaba (Qwen)CN

    Qwen3.6-35B-A3B (Qwen3.6-Flash)

  • AnthropicUS

    Claude Opus 4.7

  • PrismMLUS

    Ternary-Bonsai-1.7B · Ternary-Bonsai-4B · Ternary-Bonsai-8B

  • Tencent HunyuanCN

    POINTS-Seeker

  • VIDRAFT / FINAL-BenchKR

    Darwin-27B-KR-V3

  • Alibaba (Qwen)CN

    Qwen3.6-Max-Preview

  • Google / Google DeepMindUS

    Gemini Robotics-ER 1.6

  • NVIDIAUS

    Ising-Calibration-1-35B-A3B

  • VIDRAFT / FINAL-BenchKR

    Darwin-27B-KR-V2

  • GiniGen AIKR

    Rogue-27B-KR

  • NVIDIAUS

    Audio-Flamingo-Next-Instruct · Audio-Flamingo-Next-Think · Audio-Flamingo-Next-Captioner

  • OpenMOSS / Fudan UniversityCN

    MOSS-Audio-4B-Instruct · MOSS-Audio-4B-Thinking · MOSS-Audio-8B-Instruct · MOSS-Audio-8B-Thinking

  • VIDRAFT / FINAL-BenchKR

    Darwin-27B-KR

  • VIDRAFT / FINAL-BenchKR

    Darwin-27B-Opus

  • VIDRAFT / FINAL-BenchKR

    Darwin-4B-Genesis

  • LG AI ResearchKR

    EXAONE-4.5-33B

  • Tencent HunyuanCN

    HY-Embodied-0.5-MoT-2B

  • Trillion LabsKR

    Gravity-16B-A3B-Base

  • Liquid AIUS

    LFM2.5-VL-450M

  • Lunit ConsortiumKR

    L1-16B-A3B

  • Meta AIUS

    Muse Spark

  • OpenMOSS / Fudan UniversityCN

    MOSS-VL-Base-0408 · MOSS-VL-Instruct-0408

  • VIDRAFT / FINAL-BenchKR

    Darwin-4B-Opus

  • AppleUS

    SimpleSD-4B-instruct · SimpleSD-30B-instruct · SimpleSD-4B-thinking

  • CognitionUS

    SWE-1.6

  • NEWEN AIKR

    QuettaLLMs-27B-Koreasoner-V3

  • PoSTMEDIA AI LabKR

    Vayne-V3-Pro

  • Z.ai / Zhipu AICN

    GLM-5.1

  • VIDRAFT / FINAL-BenchKR

    Darwin-31B-Opus

  • VIDRAFT / FINAL-BenchKR

    Darwin-9B-Opus

  • AmazonUS

    BMOJOF-primed-HQwen3-8B-Instruct · GDN-primed-HQwen3-32B-Instruct · GDN-primed-HQwen3-8B-Instruct · GDN-primed-HQwen3-8B-Reasoner · GKA-primed-HQwen3-32B-Instruct · GKA-primed-HQwen3-32B-Reasoner · GKA-primed-HQwen3-8B-Instruct · GKA-primed-HQwen3-8B-Reasoner · Mamba2-primed-HQwen3-8B-Instruct

  • JD.comCN

    JoyAI-LLM-Flash · JoyAI-LLM-Flash-Base

  • Google / Google DeepMindUS

    Gemma 4 E2B Base · Gemma 4 E2B IT · Gemma 4 E4B Base · Gemma 4 E4B IT · Gemma 4 26B-A4B Base · Gemma 4 26B-A4B IT · Gemma 4 31B Base · Gemma 4 31B IT

  • Alibaba (Qwen)CN

    Qwen3.6-Plus

  • Arcee AIUS

    Trinity-Large-Thinking

  • Tencent HunyuanCN

    Sequential-Hidden-Decoding-8B-n8-Instruct

  • Z.ai / Zhipu AICN

    GLM-5V-Turbo

March 2026

68 models
LabModelsReleaseDateSource
Liquid AIUS2LFM2.5-350M, LFM2.5-350M-BaseOfficial announcement
PrismMLUS3Bonsai-1.7B, Bonsai-4B, Bonsai-8BCounted as a separately trained native low-bit model; equivalent precision/packaging conversions are excluded.Official announcement
VIDRAFT / FINAL-BenchKR1Darwin-35B-A3B-OpusGeneral text/reasoning checkpoint created by combining existing language-model weights. This is a derivative release, not evidence of a newly pretrained foundation or independently verified benchmark leadership.Model card and weightsPublisher repository history
Kuaishou (Kwaipilot)CN1KAT-Coder-V2March 27 follows the developer dated API price entry, corroborated by the March 29 public-availability statement.Official announcementDeveloper API catalog and dated pricing
NVIDIAUS1gpt-oss-puzzle-88BOfficial model card
PoSTMEDIA AI LabKR1Vayne-V3Korean enterprise conversation and tool-use fine-tune of gpt-oss-safeguard-20b; repurposed for general language generation, rather than only safety classification.Model card and weightsPublisher repository history
Allen Institute for AI (Ai2)US2MolmoWeb-4B, MolmoWeb-8BOfficial announcement
UpstageKR1Solar Pro 3 (solar-pro3-260323)Public proprietary Solar API language model; native output is text.Official model release historyOfficial model release history
FreedomIntelligence / CUHK ShenzhenCN3HuatuoGPT-3-8B, HuatuoGPT-3-32B, HuatuoGPT-3-7B-PanguDate combines dated developer release documentation with first complete-weight availability. Historical public/private visibility was not independently archived.Official announcementOfficial model cardOfficial model cardOfficial model card
MeituanCN1LongCat-Flash-ProverDate uses dated developer release documentation together with complete public-weight history. Repository history does not independently prove the exact private-to-public transition.Official announcementOfficial model card
Cursor / AnysphereUS1Composer 2Official announcement
NVIDIAUS1Nemotron-Cascade-2-30B-A3BMarch 19 model-card release date used; research page is dated March 16 and repository was created March 18.Official model card
MiniMaxCN1MiniMax M2.7Official announcement
Xiaomi MiMoCN2MiMo-V2-Pro, MiMo-V2-OmniOfficial announcement
OpenAIUS2GPT-5.4 mini, GPT-5.4 nanoOfficial announcement
NVIDIAUS1Nemotron-3-Nano-4BOfficial model card
OpenMOSS / Fudan UniversityCN2SciThinker-4B, SciThinker-30BDate combines dated developer release documentation with first complete weights; historical public/private visibility was not independently archived.Official announcementOfficial model cardOfficial model card
Z.ai / Zhipu AICN1GLM-5-TurboOfficial announcementOfficial announcement
NVIDIAUS1NVILA-8B-HD-VideoMarch 13 is the earliest verified post-upload dated developer documentation in this audit; exact public visibility could precede that date.Official announcementOfficial model card
MiroMindSG2MiroThinker-1.7-mini, MiroThinker-1.730B and 235B research-agent language checkpoints with verification-oriented post-training. H1 is a system built on these models, not an additional weight release.Model card and weightsModel card and weightsPublisherTechnical paper
NVIDIAUS2Nemotron-3-Super-120B-A12B, Nemotron-3-Super-120B-A12B-BaseMarch 11 public launch/card date used; base card has a March 4 placeholder preceding March 10 repository creation.Official model cardOfficial model cardOfficial announcement
Reka AIUS1Reka Edge 2603Official announcementOfficial model card
NVIDIAUS2Cosmos-Reason2-2B (2026-03-10), Cosmos-Reason2-8B (2026-03-10)Official model cardOfficial model card
Tencent HunyuanCN3Sequential-Hidden-Decoding-8B-n2, Sequential-Hidden-Decoding-8B-n4, Sequential-Hidden-Decoding-8B-n8Official announcement
Ant Group (inclusionAI)CN1ZwZ-2BDate uses dated developer release documentation together with complete public-weight history. Repository history does not independently prove the exact private-to-public transition.Official announcementOfficial model card
Tencent HunyuanCN2Penguin-VL-2B, Penguin-VL-8BOfficial announcement
Allen Institute for AI (Ai2)US4OLMo-Hybrid-7B, OLMo-Hybrid-Instruct-SFT-7B, OLMo-Hybrid-Instruct-DPO-7B, OLMo-Hybrid-Think-SFT-7BOfficial announcementOfficial model card
OpenAIUS2GPT-5.4, GPT-5.4 ProOfficial announcement
MicrosoftUS1Phi-4-reasoning-vision-15BOfficial announcement
Google / Google DeepMindUS1Gemini 3.1 Flash-Lite PreviewOfficial announcement
OpenAIUS1GPT-5.3 InstantOfficial announcement
Alibaba (Qwen)CN8Qwen3.5-0.8B, Qwen3.5-2B, Qwen3.5-4B, Qwen3.5-9B, Qwen3.5-0.8B-Base, Qwen3.5-2B-Base, Qwen3.5-4B-Base, Qwen3.5-9B-BaseOfficial announcementOfficial model card
IQuest LabCN8IQuest-Coder-V1-7B-Base, IQuest-Coder-V1-7B-Instruct, IQuest-Coder-V1-7B-Thinking, IQuest-Coder-V1-14B-Base, IQuest-Coder-V1-14B-Instruct, IQuest-Coder-V1-14B-Thinking, IQuest-Coder-V1-40B-Thinking, IQuest-Coder-V1-40B-Loop-ThinkingOfficial announcementOfficial announcementOfficial announcement
A*STAR (MERaLiON)SG1MERaLiON-3-10BGeneral audio-language question answering and contextual audio understanding with text output; Whisper encoder plus Gemma 2 decoder. Formerly named MERaLiON-3-10B-preview; counted once.Model card and weightsPublisherPublisher repository history
Beijing Academy of Artificial IntelligenceCN1RoboBrain2.5-4BOfficial announcementOfficial model card
  • Liquid AIUS

    LFM2.5-350M · LFM2.5-350M-Base

  • PrismMLUS

    Bonsai-1.7B · Bonsai-4B · Bonsai-8B

  • VIDRAFT / FINAL-BenchKR

    Darwin-35B-A3B-Opus

  • Kuaishou (Kwaipilot)CN

    KAT-Coder-V2

  • NVIDIAUS

    gpt-oss-puzzle-88B

  • PoSTMEDIA AI LabKR

    Vayne-V3

  • Allen Institute for AI (Ai2)US

    MolmoWeb-4B · MolmoWeb-8B

  • UpstageKR

    Solar Pro 3 (solar-pro3-260323)

  • FreedomIntelligence / CUHK ShenzhenCN

    HuatuoGPT-3-8B · HuatuoGPT-3-32B · HuatuoGPT-3-7B-Pangu

  • MeituanCN

    LongCat-Flash-Prover

  • Cursor / AnysphereUS

    Composer 2

  • NVIDIAUS

    Nemotron-Cascade-2-30B-A3B

  • MiniMaxCN

    MiniMax M2.7

  • Xiaomi MiMoCN

    MiMo-V2-Pro · MiMo-V2-Omni

  • OpenAIUS

    GPT-5.4 mini · GPT-5.4 nano

  • NVIDIAUS

    Nemotron-3-Nano-4B

  • OpenMOSS / Fudan UniversityCN

    SciThinker-4B · SciThinker-30B

  • Z.ai / Zhipu AICN

    GLM-5-Turbo

  • NVIDIAUS

    NVILA-8B-HD-Video

  • MiroMindSG

    MiroThinker-1.7-mini · MiroThinker-1.7

  • NVIDIAUS

    Nemotron-3-Super-120B-A12B · Nemotron-3-Super-120B-A12B-Base

  • Reka AIUS

    Reka Edge 2603

  • NVIDIAUS

    Cosmos-Reason2-2B (2026-03-10) · Cosmos-Reason2-8B (2026-03-10)

  • Tencent HunyuanCN

    Sequential-Hidden-Decoding-8B-n2 · Sequential-Hidden-Decoding-8B-n4 · Sequential-Hidden-Decoding-8B-n8

  • Ant Group (inclusionAI)CN

    ZwZ-2B

  • Tencent HunyuanCN

    Penguin-VL-2B · Penguin-VL-8B

  • Allen Institute for AI (Ai2)US

    OLMo-Hybrid-7B · OLMo-Hybrid-Instruct-SFT-7B · OLMo-Hybrid-Instruct-DPO-7B · OLMo-Hybrid-Think-SFT-7B

  • OpenAIUS

    GPT-5.4 · GPT-5.4 Pro

  • MicrosoftUS

    Phi-4-reasoning-vision-15B

  • Google / Google DeepMindUS

    Gemini 3.1 Flash-Lite Preview

  • OpenAIUS

    GPT-5.3 Instant

  • Alibaba (Qwen)CN

    Qwen3.5-0.8B · Qwen3.5-2B · Qwen3.5-4B · Qwen3.5-9B · Qwen3.5-0.8B-Base · Qwen3.5-2B-Base · Qwen3.5-4B-Base · Qwen3.5-9B-Base

  • IQuest LabCN

    IQuest-Coder-V1-7B-Base · IQuest-Coder-V1-7B-Instruct · IQuest-Coder-V1-7B-Thinking · IQuest-Coder-V1-14B-Base · IQuest-Coder-V1-14B-Instruct · IQuest-Coder-V1-14B-Thinking · IQuest-Coder-V1-40B-Thinking · IQuest-Coder-V1-40B-Loop-Thinking

  • A*STAR (MERaLiON)SG

    MERaLiON-3-10B

  • Beijing Academy of Artificial IntelligenceCN

    RoboBrain2.5-4B

February 2026

58 models
LabModelsReleaseDateSource
Shanghai AI LaboratoryCN1Spatial-SSRL-3BOfficial model card
Alibaba (Qwen)CN1Qwen3.5-35B-A3B-BaseBase date uses checkpoint upload history corroborating family launch; repository metadata does not expose public-visibility transition.Official model cardOfficial model card
InceptionUS1Mercury 2Official announcementOfficial announcement
Liquid AIUS1LFM2-24B-A2BOfficial announcement
NVIDIAUS3Nemotron-Terminal-8B, Nemotron-Terminal-14B, Nemotron-Terminal-32BFirst dated developer statement explicitly releasing weights is February 24; earlier repository creation is not used as a release date.Official announcementOfficial model cardOfficial model cardOfficial model card
Alibaba (Qwen)CN3Qwen3.5-35B-A3B (Qwen3.5-Flash), Qwen3.5-27B, Qwen3.5-122B-A10BOfficial announcementOfficial model card
Tech MahindraIN1Project Indus 8B8B Hindi-first education LLM built with NVIDIA NeMo under the IndiaAI Mission.Press release
Google / Google DeepMindUS1Gemini 3.1 Pro PreviewOfficial announcement
Trillion LabsKR1Tri-21B-ThinkReasoning checkpoint with reinforcement learning, separate from Think-Preview and the 2025 Tri-21B base.Model card and weightsPublisherPublisher repository history
Trillion LabsKR1Tri-21B-Think-PreviewReasoning preview post-trained from the pre-existing Tri-21B model.Model card and weightsPublisherPublisher repository history
BharatGenIN1Param2 17B-A2.4B Thinking17B MoE (2.4B active) trained from scratch across 22 Indian languages, open weights under a non-commercial license.Hugging FaceBharatGen
Sarvam AIIN2Sarvam 30B, Sarvam 105BTrained from scratch, open weights under Apache 2.0.SarvamHugging Face
Alibaba (Qwen)CN2WebWorld-14B, WebWorld-32BDate combines dated developer release documentation with complete checkpoint availability; historical public/private visibility was not independently archived.Official announcementOfficial model cardOfficial model card
AnthropicUS1Claude Sonnet 4.6Official announcement
BharatGenIN1Patram1.2 7B InstructReads Indian documents, charts and tables. Built on OLMo-7B, open weights under Apache 2.0.Hugging Face
NVIDIAUS1Nemotron-Nano-9B-v2-JapaneseOfficial model card
xAIUS1Grok 4.20 BetaOfficial release notes
Alibaba (Qwen)CN1WebWorld-8BDate combines dated developer release documentation with complete checkpoint availability; historical public/private visibility was not independently archived.Official announcementOfficial model card
Ant Group (inclusionAI)CN1Ling-2.5-1TDate uses dated developer release documentation together with complete public-weight history. Repository history does not independently prove the exact private-to-public transition.Official announcementOfficial model card
Alibaba (Qwen)CN1Qwen3.5-397B-A17B (Qwen3.5-Plus)Official announcementOfficial model card
ByteDance SeedCN4Seed2.0 Pro, Seed2.0 Lite, Seed2.0 Mini, Seed2.0 CodeOfficial announcement
Ant Group (inclusionAI)CN3ZwZ-4B, ZwZ-7B, ZwZ-8BDate uses dated developer release documentation together with complete public-weight history. Repository history does not independently prove the exact private-to-public transition.Official announcementOfficial model cardOfficial model cardOfficial model card
Nanbeige LLM Lab / BOSS ZhipinCN1Nanbeige4.1-3BFirst dated developer release statement verified is the February 13 paper; repository creation was earlier and is not treated as release.Official announcementOfficial model cardOfficial organization attribution
Ant Group (inclusionAI)CN1Ring-2.5-1TDate uses dated developer release documentation together with complete public-weight history. Repository history does not independently prove the exact private-to-public transition.Official announcementOfficial model card
MiniMaxCN1MiniMax M2.5Official announcement
OpenAIUS1GPT-5.3-Codex-SparkOfficial announcement
OpenBMB / ModelBestCN1MiniCPM-SALAOfficial announcementOfficial model card
Z.ai / Zhipu AICN1GLM-5English Z.ai announcement is February 12; provider launch date and weight history support February 11.Official announcementOfficial model cardOfficial announcement
Ant Group (inclusionAI)CN2LLaDA2.1-mini, LLaDA2.1-flashDate uses dated developer release documentation together with complete public-weight history. Repository history does not independently prove the exact private-to-public transition.Official announcementOfficial model cardOfficial model card
Ant Group (inclusionAI)CN3UI-Venus-1.5-2B, UI-Venus-1.5-8B, UI-Venus-1.5-30B-A3BDate is the first dated developer report explicitly linking released checkpoints, corroborated by same-day complete weights; exact earlier visibility is not independently logged.Official announcementOfficial model cardOfficial model cardOfficial model card
Cursor / AnysphereUS1Composer 1.5Official announcement
iFLYTEKCN1Spark X2February 9 is the availability date documented by the official HTTP and WebSocket guides; wider launch coverage appeared February 11.Official announcementOfficial announcement
AI SingaporeSG2Apertus-SEA-LION-v4-8B-IT, Gemma-SEA-LION-v4-4B-VLRegional instruction tuning of Apertus 8B and Gemma 3 4B; the Gemma variant accepts images and produces text.Model card and weightsModel card and weightsPublisher
AnthropicUS1Claude Opus 4.6Official announcement
MeituanCN1LongCat-Flash-LiteUses documented February 5 public API release; January weight upload and February 2 tracker dates do not independently establish first public access.Official announcement
OpenAIUS1GPT-5.3-CodexOfficial announcement
Shanghai AI LaboratoryCN1Intern-S1-ProDate is the first dated developer release documentation corroborated by complete model weights; historical repository visibility was not independently archived.Official announcementOfficial model card
Xiaomi MiMoCN1MiMo-V2-Flash-0204Official announcement
Allen Institute for AI (Ai2)US1SERA-14BOfficial announcement
MicrosoftUS1RE-TRAC-30B-A3BPublic demo February 3; downloadable weights followed February 5 under the InfoAgent repository name.Official announcementOfficial model card
Alibaba (Qwen)CN2Qwen3-Coder-Next, Qwen3-Coder-Next-BaseOfficial blog date February 2; technical-report accessed field February 3.Official announcementOfficial model cardOfficial model card
StepFunCN1Step-3.5-FlashOfficial announcement
  • Shanghai AI LaboratoryCN

    Spatial-SSRL-3B

  • Alibaba (Qwen)CN

    Qwen3.5-35B-A3B-Base

  • InceptionUS

    Mercury 2

  • Liquid AIUS

    LFM2-24B-A2B

  • NVIDIAUS

    Nemotron-Terminal-8B · Nemotron-Terminal-14B · Nemotron-Terminal-32B

  • Alibaba (Qwen)CN

    Qwen3.5-35B-A3B (Qwen3.5-Flash) · Qwen3.5-27B · Qwen3.5-122B-A10B

  • Tech MahindraIN

    Project Indus 8B

  • Google / Google DeepMindUS

    Gemini 3.1 Pro Preview

  • Trillion LabsKR

    Tri-21B-Think

  • Trillion LabsKR

    Tri-21B-Think-Preview

  • BharatGenIN

    Param2 17B-A2.4B Thinking

  • Sarvam AIIN

    Sarvam 30B · Sarvam 105B

  • Alibaba (Qwen)CN

    WebWorld-14B · WebWorld-32B

  • AnthropicUS

    Claude Sonnet 4.6

  • BharatGenIN

    Patram1.2 7B Instruct

  • NVIDIAUS

    Nemotron-Nano-9B-v2-Japanese

  • xAIUS

    Grok 4.20 Beta

  • Alibaba (Qwen)CN

    WebWorld-8B

  • Ant Group (inclusionAI)CN

    Ling-2.5-1T

  • Alibaba (Qwen)CN

    Qwen3.5-397B-A17B (Qwen3.5-Plus)

  • ByteDance SeedCN

    Seed2.0 Pro · Seed2.0 Lite · Seed2.0 Mini · Seed2.0 Code

  • Ant Group (inclusionAI)CN

    ZwZ-4B · ZwZ-7B · ZwZ-8B

  • Nanbeige LLM Lab / BOSS ZhipinCN

    Nanbeige4.1-3B

  • Ant Group (inclusionAI)CN

    Ring-2.5-1T

  • MiniMaxCN

    MiniMax M2.5

  • OpenAIUS

    GPT-5.3-Codex-Spark

  • OpenBMB / ModelBestCN

    MiniCPM-SALA

  • Z.ai / Zhipu AICN

    GLM-5

  • Ant Group (inclusionAI)CN

    LLaDA2.1-mini · LLaDA2.1-flash

  • Ant Group (inclusionAI)CN

    UI-Venus-1.5-2B · UI-Venus-1.5-8B · UI-Venus-1.5-30B-A3B

  • Cursor / AnysphereUS

    Composer 1.5

  • iFLYTEKCN

    Spark X2

  • AI SingaporeSG

    Apertus-SEA-LION-v4-8B-IT · Gemma-SEA-LION-v4-4B-VL

  • AnthropicUS

    Claude Opus 4.6

  • MeituanCN

    LongCat-Flash-Lite

  • OpenAIUS

    GPT-5.3-Codex

  • Shanghai AI LaboratoryCN

    Intern-S1-Pro

  • Xiaomi MiMoCN

    MiMo-V2-Flash-0204

  • Allen Institute for AI (Ai2)US

    SERA-14B

  • MicrosoftUS

    RE-TRAC-30B-A3B

  • Alibaba (Qwen)CN

    Qwen3-Coder-Next · Qwen3-Coder-Next-Base

  • StepFunCN

    Step-3.5-Flash

January 2026

52 models
LabModelsReleaseDateSource
China Telecom / TeleAICN1TeleChat3-Coder-36B-ThinkingOfficial model card
Trillion LabsKR1gWorld-32BQwen3-VL adaptation that predicts GUI state as reasoning plus executable HTML/CSS. Native output is code/text, not pixels or robot trajectories.Model card and weightsPublisherTechnical paperPublisher repository history
Trillion LabsKR1gWorld-8BQwen3-VL adaptation that predicts GUI state as reasoning plus executable HTML/CSS. Native output is code/text, not pixels or robot trajectories.Model card and weightsPublisherTechnical paperPublisher repository history
OpenBMB / ModelBestCN3NOSA-1B, NOSA-3B, NOSA-8BDate uses dated developer release documentation together with complete public-weight history. Repository history does not independently prove the exact private-to-public transition.Official announcementOfficial model cardOfficial model cardOfficial model card
Allen Institute for AI (Ai2)US4SERA-8B, SERA-32B, SERA-8B-GA, SERA-32B-GAOfficial announcementOfficial model card
Arcee AIUS3Trinity-Large-Preview, Trinity-Large-Base, Trinity-Large-TrueBaseOfficial announcement
Moonshot AICN1Kimi K2.5Official announcementOfficial announcementOfficial announcement
Tencent HunyuanCN1Youtu-VL-4B-InstructDate uses dated developer release documentation together with complete public-weight history. Repository history does not independently prove the exact private-to-public transition.Official announcementOfficial model card
UpstageKR1Solar Pro 3 (solar-pro3-260126)Public proprietary Solar API language model; native output is text.Official model release historyOfficial model release history
Alibaba (Qwen)CN1Qwen3-Max-Thinking (2026-01-23)Official announcement
Alibaba (Qwen)CN1Qwen3-VL-Flash-2026-01-22Official announcement
ByteDance SeedCN2Stable-DiffCoder-8B-Base, Stable-DiffCoder-8B-InstructOfficial announcement
KakaoKR1kanana-2-30b-a3b-base-2601Updated January base checkpoint in Kakao’s 30B/3B-active family.Model card and weightsPublisher
KakaoKR2kanana-2-30b-a3b-instruct-2601, kanana-2-30b-a3b-thinking-2601Instruction-following and thinking variants of the January Kanana-2 MoE family.Model card and weightsModel card and weightsPublisher
KakaoKR1kanana-2-30b-a3b-mid-2601January mid-training checkpoint with reasoning-oriented training.Model card and weightsPublisher
Liquid AIUS1LFM2.5-1.2B-ThinkingOfficial announcement
OpenBMB / ModelBestCN1AgentCPM-ReportOfficial announcement
OpenDataLabCN4ChartVerse-2B, ChartVerse-4B, ChartVerse-8B, ChartVerse-CoderUses the dated developer report explicitly releasing checkpoints after complete weights; exact earlier public visibility is not independently logged.Official announcementOfficial announcementOfficial model cardOfficial model cardOfficial model cardOfficial model card
NVIDIAUS1Qwen3-8B-DMS-8xOfficial model card
Trillion LabsKR1Trida-7BGeneral Korean/English/Japanese block-diffusion language model continued from Tri-7B; distinct from the 2025 preview.Model card and weightsPublisher repository history
Z.ai / Zhipu AICN1GLM-4.7-FlashOfficial release notesOfficial model card
StepFunCN2Step3-VL-10B, Step3-VL-10B-BaseJanuary 15 is the dated public project launch; earlier January 13 weight commits and January 14 paper submission do not independently establish public visibility.Official announcementOfficial announcementOfficial announcementOfficial model card
MeituanCN1LongCat-Flash-Thinking-2601Official announcement
Alibaba (Qwen)CN1Qwen-Flash-CharacterOfficial announcement
Google / Google DeepMindUS1MedGemma 1.5 4B ITOfficial announcement
MicrosoftUS2FrogBoss-32B-2510, FrogMini-14B-2510Official model cardOfficial model card
OpenBMB / ModelBestCN1AgentCPM-ExploreOfficial announcement
Xiaomi MiMoCN1MiMo-V2-Flash-0112Official announcement
Beijing Academy of Artificial IntelligenceCN2RoboBrain2.5-8B-NV, RoboBrain2.5-8B-MTOfficial announcementOfficial model cardOfficial model card
Nous ResearchUS1NousCoder-14BOfficial post renders January 7 at 01:09 in the India-time research browser (January 6 UTC). The UTC calendar date is used consistently with timestamp-only sources.Official announcementOfficial model card
Liquid AIUS4LFM2.5-1.2B-Base, LFM2.5-1.2B-Instruct, LFM2.5-1.2B-JP, LFM2.5-VL-1.6BOfficial announcement
MiroMindSG2MiroThinker-v1.5-30B, MiroThinker-v1.5-235BTool-using research agents post-trained from Qwen3 Thinking checkpoints; text answers and tool calls.Model card and weightsModel card and weightsPublisher
SK TelecomKR1A.X-K1519B/33B-active general language and reasoning model, trained by the SKT-led consortium. Announced at the December 2025 presentation; downloadable artifacts cross the 2026 boundary in Korean time.Model card and weightsPublisherPublisherPublisher repository history
  • China Telecom / TeleAICN

    TeleChat3-Coder-36B-Thinking

  • Trillion LabsKR

    gWorld-32B

  • Trillion LabsKR

    gWorld-8B

  • OpenBMB / ModelBestCN

    NOSA-1B · NOSA-3B · NOSA-8B

  • Allen Institute for AI (Ai2)US

    SERA-8B · SERA-32B · SERA-8B-GA · SERA-32B-GA

  • Arcee AIUS

    Trinity-Large-Preview · Trinity-Large-Base · Trinity-Large-TrueBase

  • Moonshot AICN

    Kimi K2.5

  • Tencent HunyuanCN

    Youtu-VL-4B-Instruct

  • UpstageKR

    Solar Pro 3 (solar-pro3-260126)

  • Alibaba (Qwen)CN

    Qwen3-Max-Thinking (2026-01-23)

  • Alibaba (Qwen)CN

    Qwen3-VL-Flash-2026-01-22

  • ByteDance SeedCN

    Stable-DiffCoder-8B-Base · Stable-DiffCoder-8B-Instruct

  • KakaoKR

    kanana-2-30b-a3b-base-2601

  • KakaoKR

    kanana-2-30b-a3b-instruct-2601 · kanana-2-30b-a3b-thinking-2601

  • KakaoKR

    kanana-2-30b-a3b-mid-2601

  • Liquid AIUS

    LFM2.5-1.2B-Thinking

  • OpenBMB / ModelBestCN

    AgentCPM-Report

  • OpenDataLabCN

    ChartVerse-2B · ChartVerse-4B · ChartVerse-8B · ChartVerse-Coder

  • NVIDIAUS

    Qwen3-8B-DMS-8x

  • Trillion LabsKR

    Trida-7B

  • Z.ai / Zhipu AICN

    GLM-4.7-Flash

  • StepFunCN

    Step3-VL-10B · Step3-VL-10B-Base

  • MeituanCN

    LongCat-Flash-Thinking-2601

  • Alibaba (Qwen)CN

    Qwen-Flash-Character

  • Google / Google DeepMindUS

    MedGemma 1.5 4B IT

  • MicrosoftUS

    FrogBoss-32B-2510 · FrogMini-14B-2510

  • OpenBMB / ModelBestCN

    AgentCPM-Explore

  • Xiaomi MiMoCN

    MiMo-V2-Flash-0112

  • Beijing Academy of Artificial IntelligenceCN

    RoboBrain2.5-8B-NV · RoboBrain2.5-8B-MT

  • Nous ResearchUS

    NousCoder-14B

  • Liquid AIUS

    LFM2.5-1.2B-Base · LFM2.5-1.2B-Instruct · LFM2.5-1.2B-JP · LFM2.5-VL-1.6B

  • MiroMindSG

    MiroThinker-v1.5-30B · MiroThinker-v1.5-235B

  • SK TelecomKR

    A.X-K1