JIUFENG
ENCHINESE

AI Model Releases

Every model release from our twice-daily AI news in one table: what, who and when, each linked to the full item and sources.

Model changes only: new models, open weights, API access, first platform listings and leaderboard ranks. No product features, pricing or single benchmark scores.

291 releasesLast updated: November 4, 2026

November 2026

  1. Apollo WorldBaiduNew model
    Baidu unveils Apollo World, a world model already used to test Apollo Go
  2. Drex Max v1Nace AIOpen weights
    Nace AI opens Drex Max v1, a 27B decision model that runs on a Mac
  3. H3 FlexPikaOpen weights
    Pika open-sources H3 Flex on MiniMax H3: joint audio-video in six steps
  4. HunyuanTencentPlatform
    WeChat Pay store robot built on Tencent Hunyuan starts work at a 7-Eleven

October 2026

  1. QwenAlibabaPlatform
    Apple Intelligence gets a China date, with Qwen handling Siri's complex tasksQwen timeline →
  2. Mistral Large 4Mistral AIOpen weights
    Mistral opens Large 4 weights: about 1T parameters, 52B active
  3. GrokxAIPlatform
    Tesla Robotaxi runs until 2 a.m. and adds Grok to Model YGrok timeline →
  4. Open-H3-Regenerate-2KNVIDIAOpen weights
    Nvidia team opens a 4K video regeneration stack built on MiniMax H3
  5. IBM Granite Switch 4.1.1IBMNew model
    IBM Granite Switch 4.1.1 patch updates speech recognition and the Decision adapter
  6. Gemini 4 CarbonGoogleNew model
    Google launches Gemini 4 Carbon with restricted cyber accessGemini timeline →
  7. DeepSeek V4.1-FlashDeepSeekLeaderboard
    DeepSeek V5 placeholder IDs appear in Harness test codeDeepSeek timeline →
  8. Apex-35B-V1.0Apex IntelligenceNew model
    Apex-35B-V1.0 builds on Qwen3.6-35B-A3B with help from AI research agents
  9. Nemotron 3.5 TranscribeNVIDIANew model
    NVIDIA releases Nemotron 3.5 Transcribe and opens the Granary v2 data pipeline
  10. Kimi 3.1 ProMoonshot AIPlatform
    Doubao Work opens Kimi 3.1 Pro, GLM 5.3 and DeepSeek V4.1 Flash to enterprisesKimi timeline →
  11. GLM 5.3Zhipu AIPlatform
    Doubao Work opens Kimi 3.1 Pro, GLM 5.3 and DeepSeek V4.1 Flash to enterprisesGLM timeline →
  12. DeepSeek V4.1 FlashDeepSeekPlatform
    Doubao Work opens Kimi 3.1 Pro, GLM 5.3 and DeepSeek V4.1 Flash to enterprisesDeepSeek timeline →
  13. Ling-3.1-flash-1019Ant GroupOpen weights
    Ant Group's Bailing open-sources Ling-3.1-flash-1019, scoring 44 on the Intelligence Index
  14. Genie 4GoogleNew model
    Genie 4 world model generates sound and character dialogue for the first time
  15. GPT-6 AstraOpenAILeaderboard
    GPT-6 Astra ranks first on Retell's voice agent benchmarkOpenAI timeline →
  16. LFM3Liquid AIOpen weights
    Liquid AI releases LFM3: six open-weight 2B-class multimodal models
  17. GPT-6.1 TerraOpenAINew model
    GPT-6.1 Terra launches for computer use at half the standard API priceOpenAI timeline →
  18. SparkAudio 1.0iFlytekNew model
    iFLYTEK details SparkAudio 1.0, a 30B-A3B speech model trained on domestic chips
  19. Raon-1 miniKRAFTONOpen weights
    KRAFTON previews Raon-1 mini: a 30B-A4B open model under Apache 2.0
  20. Nemotron 3.5 Bolt E4B 4GBNVIDIAOpen weights
    NVIDIA releases Nemotron 3.5 Bolt E4B 4GB, a small open model that runs in about 4GB of VRAM
  21. Kimi K3.1 ProMoonshot AIPlatform
    Kimi K3.1 Pro rolls out in Kimi Code, with a faster Highspeed tierKimi timeline →
  22. DessaBlockNew model
    Block unveils Dessa, a model trained from scratch on human conversation
  23. Kimi K3Moonshot AIPlatform
    GitLab 19.4 adds hosted Kimi K3, MiniMax M3 and GLM 5.3Kimi timeline →
  24. MiniMax M3MiniMaxPlatform
    GitLab 19.4 adds hosted Kimi K3, MiniMax M3 and GLM 5.3MiniMax timeline →
  25. GLM 5.3Zhipu AIPlatform
    GitLab 19.4 adds hosted Kimi K3, MiniMax M3 and GLM 5.3GLM timeline →
  26. M3.1 FlashMiniMaxOpen weights
    MiniMax opens M3.1 Flash weights, with day-0 vLLM support
  27. Granite Speech 5.0 2BIBMNew model
    IBM releases Granite Speech 5.0 2B, #2 on the Open ASR Leaderboard
  28. Granite Speech 5.0 2BIBMLeaderboard
    IBM releases Granite Speech 5.0 2B, #2 on the Open ASR Leaderboard
  29. Kumo RelationalNVIDIANew model
    NVIDIA Kumo Relational beats supervised ML on 41 relational tasks with no training
  30. Voxtral 3Mistral AINew model
    Mistral ships Voxtral 3: transcription, TTS and a voice agent SDK
  31. Eleven RealtimeElevenLabsNew model
    ElevenLabs' Eleven Realtime scores 98.8% on Big Bench Audio at 0.57s
  32. Laplace 3Bi14New model
    i14 trains linear-attention Laplace 3B on Trainium2 with a fixed 29.5 MB state
  33. EdenLiveVASTOpen weights
    VAST open-sources EdenLive, a real-time world model that runs 720P at 24+ FPS on one RTX 5090
  34. Mercury 3InceptionNew model
    Mercury 3 diffusion LLM hits 1,068 tok/s and leads Haiku 5.5 on 8 of 12 self-reported tests
  35. MiMo-V2.6-ProXiaomiPlatform
    Xiaomi MiMo-V2.6-Pro launches on WorkBuddy
  36. DeepSeek V4.1-FlashDeepSeekLeaderboard
    DeepSeek V4.1-Flash update tops Arena's open agent rankingsDeepSeek timeline →
  37. MiniMax M3.1 FlashMiniMaxNew model
    MiniMax launches M3.1 Flash: 428B parameters, Intelligence Index 42MiniMax timeline →
  38. Kling 4.0KuaishouNew model
    Kling 4.0 generates 30-second clips and reads 8,000-token prompts
  39. HybridAgentTogether AIOpen weights
    togethercomputer open-sources HybridAgent, which reasons in latent space and acts through text
  40. StartLux-Decision-v1.1-27B InstantStartLuxLeaderboard
    StartLux-Decision-v1.1-27B Instant tops Decision Index 0.3.2
  41. REKIKyutaiOpen weights
    Kyutai releases REKI, a 2.4B robot VLA pretrained only on public data
  42. GLM-LegalZhipu AILeaderboard
    Zhipu's GLM-Legal leads the open-weight rankings on two Vals AI legal benchmarksGLM timeline →
  43. DeepSeek V4.1-Flash-1017DeepSeekOpen weights
    DeepSeek open-sources V4.1-Flash-1017 and opens Search and Fetch APIsDeepSeek timeline →
  44. Reflection BeamReflection AIPlatform
    Reflection Beam arrives on SageMaker as an open-weight model you can self-host
  45. Atria DawnShanghai AI LabNew model
    Shanghai AI Lab releases the full Atria Dawn agent model with 512K context
  46. Step 5 PreviewStepFunOpen weights
    StepFun opens Step 5 Preview weights: 600B MoE, 1M context
  47. d1-ultraLiquid AINew model
    Liquid AI releases d1-ultra decision model, scores 65 on the AA Decision Index
  48. Muse Spark OpenMetaOpen weights
    Meta releases Muse Spark Open weights: an 816B MoE with 1M context
  49. GPT-Live-1.1OpenAINew model
    GPT-Live-1.1 adds voices in six languages across ChatGPT, Codex and Dots
  50. TimestralMistral AINew model
    Mistral launches Timestral, its first time-series foundation models, up to 209M parameters
  51. Nemotron 3.5 Super VLNVIDIAOpen weights
    NVIDIA releases Nemotron 3.5 Super VL, a 120B open multimodal model
  52. Granite Switch 4.1IBMNew model
    IBM Granite Switch 4.1 handles chat, decisions and speech-to-text from one endpoint
  53. Jev 1.14TypeSafe AINew model
    Jev 1.14 adds image input and an optional 262k context
  54. Grok Voice 3.0xAILeaderboard
    Grok Voice 3.0 tops the speech-to-speech index, still $0.08 a minuteGrok timeline →
  55. HiRO-ACEv2Ai2New model
    Ai2's HiRO-ACEv2 generates global 3 km climate data at about 1/300th the cost of X-SHiELD
  56. Gemini 4 ArgonGoogleAPI
    Gemini 4 Argon goes GA with up to 1M output tokensGemini timeline →
  57. Muse WonderMetaNew model
    Meta releases two Muse Wonder models, with the Full version scoring 56
  58. Gemini 3.5 Live TranslateGoogleNew model
    Gemini 3.5 Live Translate covers 70+ languagesGemini timeline →
  59. GLM-5.3-HighspeedZhipu AINew model
    Zhipu's GLM-5.3-Highspeed in gray testing, reportedly ~490 tokens/sGLM timeline →
  60. Microsoft-Decision-1MicrosoftNew model
    Microsoft releases Microsoft-Decision-1, built on Qwen3.5-9B
  61. Step 5 PreviewStepFunLeaderboard
    StepFun's Step 5 Preview tops OpenRouter Trending, with open weights due October 15
  62. Qwen-Image-2.1-TurboAlibabaOpen weights
    Qwen-Image-2.1-Turbo ships open weights with 8-step generation and editingQwen timeline →
  63. GPT-6 LunaOpenAIAPI
    OpenAI Decisions API enters public beta on GPT-6 LunaOpenAI timeline →
  64. Doubao 2.1 LiteByteDancePlatform
    Doubao 2.1 Lite comes to Doubao Work, along with a canvas mode
  65. Seedream 5.0 FlashByteDancePlatform
    Doubao 2.1 Lite comes to Doubao Work, along with a canvas mode
  66. BeamReflection AIOpen weights
    Reflection unveils Beam, a 501B-parameter open-weight MoE model
  67. Claude Haiku 5.5AnthropicLeaderboard
    Claude Haiku 5.5 jumps to 72.4% on OSWorld and tops Artificial Analysis' small-model tierClaude timeline →
  68. GPT-6OpenAIPlatform
    GPT-6 rolls out in ChatGPT; OpenAI says Codex and Work have 40M active usersOpenAI timeline →
  69. Claude Haiku 5.5AnthropicNew model
    Claude Haiku 5.5 launches at about 75% lower average running cost than Haiku 4.5Claude timeline →
  70. d1-3BLiquid AIOpen weights
    Liquid AI releases open weights for d1-3B and d1-omni-600M multimodal decision models
  71. d1-omni-600MLiquid AIOpen weights
    Liquid AI releases open weights for d1-3B and d1-omni-600M multimodal decision models
  72. Mistral Large 4Mistral AINew model
    Mistral Large 4 enters public preview as a 1.05T MoE with 1M context
  73. PolicyLM-1.7BMusubiOpen weights
    Musubi releases PolicyLM-1.7B, an open-weight moderation model
  74. Intern-DecisionShanghai AI LabOpen weights
    Shanghai AI Lab open-sources Intern-Decision small decision models
  75. GLM-5.3Zhipu AIPlatform
    GLM-5.3 arrives on Amazon Bedrock for coding and long-horizon agentsGLM timeline →
  76. JEPA-AnythingPhAI LabsNew model
    JEPA-Anything: one world model across seven fields, from physics to biology
  77. Qwen3.8-maxAlibabaLeaderboard
    Qwen3.8-max tops the VA-Bench robot-arm benchmark at only 53.93%Qwen timeline →
  78. GPT-6.1 Sol ProOpenAILeaderboard
    GPT-6.1 Sol Pro leads newsroom benchmarkOpenAI timeline →
  79. KolibriAleph AlphaOpen weights
    Kolibri releases 78GB FP8 open weights
  80. Strands Decider 2BAmazonOpen weights
    Strands Decider 2B targets local decisions
  81. Utopai XUtopai StudiosLeaderboard
    Utopai X ranks second on audio-video leaderboard
  82. AstaBriefAi2Open weights
    AllenAI open-sources AstaBrief report model
  83. ClefCloudflareOpen weights
    Cloudflare releases Qwen-based Clef decision models
  84. Clef-flashCloudflareOpen weights
    Cloudflare releases Qwen-based Clef decision models
  85. Claude Opus 5.5AnthropicLeaderboard
    Claude Opus 5.5 leads APEX AccountingClaude timeline →
  86. Qwen3.8-27BAlibabaPlatform
    Qwen3.8-27B arrives on Nebius Token FactoryQwen timeline →
  87. GPT-6 Astra UltrafastOpenAIAPI
    GPT-6 Astra Ultrafast offers up to 8× token generation speedOpenAI timeline →
  88. GriffinTavusNew model
    Tavus introduces Griffin with a limited research preview
  89. Gemini 4 ArgonGoogleNew model
    Gemini 4 Argon launches for select cyber defendersGemini timeline →

September 2026

  1. GPT-6.1 SolOpenAINew model
    GPT-6.1 Sol debuts at DevDay with $2/$10 API pricingOpenAI timeline →
  2. Mercury VoiceInceptionNew model
    Inception's Mercury Voice diffusion model posts a 320ms median time to first token
  3. d1Liquid AINew model
    Liquid AI's d1 returns probabilities with zero output tokens
  4. Kumo TabularNVIDIAOpen weights
    NVIDIA opens Kumo Tabular for training-free table prediction
  5. Qwen-Audio-3.1-RealtimeAlibabaNew model
    Qwen-Audio-3.1-Realtime launches with an ~85% price cutQwen timeline →
  6. Claude Sonnet 5.5AnthropicNew model
    Claude Sonnet 5.5 lands with up to 30% lower per-task costClaude timeline →
  7. Grok 4.7xAIPlatform
    Grok 4.7 arrives on Amazon Bedrock with a 500K context windowGrok timeline →
  8. Qwen3-TTSAlibabaPlatform
    Qwen3-TTS on SageMaker, streaming speech before generation endsQwen timeline →
  9. Qwen-Image-2512AlibabaOpen weights
    Qwen-Image-2512 Released, Billed as Top Open Image ModelQwen timeline →
  10. Holo4H CompanyOpen weights
    H Ships Holo4 in 27B Dense and 35B-A3B MoE
  11. Holotron4 NanoH CompanyNew model
    H Ships Holo4 in 27B Dense and 35B-A3B MoE
  12. Ember-1FireworksNew model
    Ember-1 trims Kimi K3 reasoning tokens by about 40%
  13. Space BunnyUndisclosedLeaderboard
    Anonymous model Space Bunny tops OpenRouter's daily call chart
  14. FermiQLLM 1.0费米宇宙New model
    FermiQLLM 1.0 adds quantum methods to a Qwen base
  15. Saaras V4SarvamNew model
    Sarvam's Saaras V4 covers all 22 official Indian languages
  16. MiniMax M3.1-Flash-PreviewMiniMaxNew model
    MiniMax ships M3.1-Flash-Preview on MiniMax CodeMiniMax timeline →
  17. LFM2.5-VL-3B-DSparkLiquid AIOpen weights
    Liquid releases a VLM drafter with up to 3.13x decoding
  18. AWETARS RoboticsNew model
    它石智航 (Tashi Zhihang) details an AWE embodied base model and three hardware lines
  19. Opus 5.5AnthropicNew model
    Opus 5.5 lands at $20 per million output tokens
  20. GPT-6 SolOpenAINew model
    GPT-6 Sol and Luna arrive at half the previous token priceOpenAI timeline →
  21. GPT-6 LunaOpenAINew model
    GPT-6 Sol and Luna arrive at half the previous token priceOpenAI timeline →
  22. Claude Opus 5.5AnthropicPlatform
    Claude Opus 5.5 lands on Amazon Bedrock at a claimed 40% lower costClaude timeline →
  23. MiMo-V2.6-ProXiaomiLeaderboard
    MiMo-V2.6-Pro tops open model rankings at 46 points
  24. Grok 4.7xAINew model
    Grok 4.7 ships at Grok 4.6's $2/$6 token pricesGrok timeline →
  25. Qwen-Image-2.1AlibabaOpen weights
    Qwen-Image-2.1 outputs transparent images natively, commercial use needs a separate licenseQwen timeline →
  26. Qwen3.8-LiveTranslateAlibabaNew model
    Qwen3.8-LiveTranslate cuts average interpretation lag to 2.3 secondsQwen timeline →
  27. Step 5 PreviewStepFunLeaderboard
    StepFun's Step 5 Preview enters the open-weight top three at one-eighth of Opus 5's task cost
  28. Qwen3.8-Omni-FlashAlibabaNew model
    Qwen3.8-Omni-Flash prices inputs at a fifth of Gemini FlashQwen timeline →
  29. Xing4.0-29B-A4BChina TelecomOpen weights
    China Telecom open-sources Xing4.0-29B-A4B
  30. jina-ocr-v1Jina AIOpen weights
    jina-ocr-v1 open weights: 3.4B total, 570M active per token
  31. Kimi K3Moonshot AIPlatform
    Kimi K3 lands on Amazon Bedrock with 2.8T parametersKimi timeline →
  32. Grok Voice Transcribe 2.0xAINew model
    Grok Voice Transcribe 2.0 ships, Loom transcribes every video with itGrok timeline →
  33. LimiX-2Tsinghua & Stable AINew model
    Tsinghua and Stable AI release LimiX-2 for structured data
  34. ZDTaichu5.0-9BZidong TaichuOpen weights
    ZDTaichu5.0-9B opens weights, claims 8 firsts on spatial tests
  35. JevTypeSafe AINew model
    TypeSafe exits stealth with $40M and Jev, a model that outputs decisions
  36. Mistral Small 4Mistral AIPlatform
    Firefox's AI assistant now runs on Mistral Small 4
  37. KoaSalesforceNew model
    Salesforce's first CRM reasoning model is a post-trained Nemotron 3 Super
  38. Doubao 2.1 ProByteDanceNew model
    Doubao 2.1 Pro ships an 0915 build, wired into Doubao Work
  39. Gemini 3.8 LiveGoogleNew model
    Gemini 3.8 Live ships with background tool callsGemini timeline →
  40. GeminiGooglePlatform
    Apple's rebuilt Siri runs on Google's Gemini, but not in the EUGemini timeline →
  41. StepAudio 3StepFunNew model
    StepFun ships five StepAudio 3 models, claiming top rankings
  42. OM-1Reward AINew model
    Reward AI releases OM-1, trained only on human glove demonstrations
  43. GPT-6 Astra MaxOpenAILeaderboard
    Arena's two leaderboards have different winnersOpenAI timeline →
  44. Claude Fable 5.1 MAnthropicLeaderboard
    Arena's two leaderboards have different winnersClaude timeline →
  45. Kimi K2.8 PreviewMoonshot AIPlatform
    Kimi K2.8 Preview is fully rolled out to Kimi Code and WorkKimi timeline →
  46. GPT-6 AstraOpenAILeaderboard
    GPT-6 Astra tops two agent benchmarks and sweeps all five drone tasksOpenAI timeline →
  47. SWE-2CognitionNew model
    Cognition releases SWE-2, post-trained from Kimi K3 at 64% lower cost
  48. Motus2ShengshuNew model
    Shengshu's Motus2 folds prediction and scoring into an action model (Chinese-language source)
  49. Fugu MaxSakana AINew model
    Sakana AI launches Fugu Max and Fugu Ultra v2 for learned multi-model orchestration
  50. Fugu Ultra v2Sakana AINew model
    Sakana AI launches Fugu Max and Fugu Ultra v2 for learned multi-model orchestration
  51. Marengo Embed 3.0TwelveLabsPlatform
    Bedrock Knowledge Bases adds Marengo Embed 3.0 for semantic video search
  52. DeepSeek-V4.1-FlashDeepSeekOpen weights
    DeepSeek-V4.1-Flash reaches GA with an 890-byte-per-token KV cacheDeepSeek timeline →
  53. Qwen 2.4TAlibabaOpen weights
    Qwen opens a 2.4T model to self-hosting with 95B active paramsQwen timeline →
  54. Suno v6SunoNew model
    Suno rebuilds its training data and ships v6 in three tiers
  55. PatchTST-FM-r2IBMNew model
    IBM ships a new Granite time-series foundation model
  56. V4.1 FlashDeepSeekNew model
    V4.1 Flash enters a time-boxed internal beta on a natively multimodal architecture
  57. GPT-6 AstraOpenAIPlatform
    GPT-6 Astra is generally available on Amazon BedrockOpenAI timeline →
  58. Ling-3.0-flash-FinAnt GroupOpen weights
    Ant's Ling team open-sources a finance-tuned model and a matching research benchmark
  59. GPT Image 2.5OpenAINew model
    GPT Image 2.5 ships, with Flare and Sunburst splitting the APIOpenAI timeline →
  60. Ling-3.0-flash-VLAnt GroupOpen weights
    Ant open-sources Ling-3.0-flash-VL: 124B total, 5.5B active
  61. Mercury 2.5InceptionNew model
    Mercury 2.5 lands on OpenRouter at a claimed 1,107 tokens per second
  62. Mercury 2.5InceptionPlatform
    Mercury 2.5 lands on OpenRouter at a claimed 1,107 tokens per second
  63. NeoHorse-1基元律动New model
    Wang Yunhe's team ships NeoHorse-1 in 4B and 9B sizes (Chinese-language source)
  64. N2.5Nex-AGINew model
    Nex-AGI ships N2.5 agent family, Pro weights still pending
  65. Qwen3.8-27BAlibabaPlatform
    Perplexity's local agent now runs on Qwen3.8-27BQwen timeline →
  66. Qwen-Drive 1.0AlibabaNew model
    Qwen-Drive 1.0 merges driving system and cockpit assistantQwen timeline →
  67. Hy4 previewTencentNew model
    Tencent Hy4 preview update cuts token use in agent tasks
  68. HUMAIN-M3HUMAINNew model
    HUMAIN-M3 builds Saudi Arabic model on MiniMax M3
  69. GPT-6 AstraOpenAIAPI
    GPT-6 Astra opens to Pro, enterprise and API usersOpenAI timeline →
  70. GPT-6 AstraOpenAIPlatform
    Replit runs GPT-6 Astra one day after releaseOpenAI timeline →
  71. Lyria 3.5GoogleNew model
    Lyria 3.5 lands in Gemini with multi-minute songs
  72. WeatherNext 3GoogleNew model
    WeatherNext 3 adds hourly refresh and ships into Search and Gemini
  73. NeoMMENeoMMEOpen weights
    NeoMME releases 260M and 800M multimodal multilingual encoders
  74. WorldGenDeemosNew model
    Hyper3D's WorldGen builds editable 3D scenes, not single assets
  75. Gemini 3.8 FlashGoogleNew model
    Gemini 3.8 Flash arrives: third Flash in six weeks, same price as 3.7Gemini timeline →
  76. Gemini 3.8 Flash CyberGoogleNew model
    Gemini 3.8 Flash Cyber opens to defenders via the Fairwind ProgramGemini timeline →
  77. Muse Spark 1.3MetaNew model
    Meta releases Muse Spark 1.3 for long-running agents and coding (Chinese-language source)
  78. Quasar 438BMultiverse ComputingNew model
    Multiverse's Quasar 438B tops European models on the Artificial Analysis index (Chinese-language source)
  79. Quasar 438BMultiverse ComputingLeaderboard
    Multiverse's Quasar 438B tops European models on the Artificial Analysis index (Chinese-language source)
  80. Claude Fable 5.1AnthropicPlatform
    Claude Fable 5.1 lands on AWS as a "Covered Model"Claude timeline →
  81. OrbisViskoNew model
    Visko launches Orbis, a "Live Model" for continuously running video
  82. ZevaTsinghua AIRNew model
    Tsinghua AIR's self-evolving Zeva lifts success rate from 26% to 73%
  83. DeepSeek-V4-Flash-Vision-ExpDeepSeekOpen weights
    DeepSeek open-sources first V4 vision model, 305B under MITDeepSeek timeline →
  84. GPT-5.6 SolOpenAIPlatform
    Cerebras serves GPT-5.6 Sol at 750 tokens/second, unchanged weightsOpenAI timeline →
  85. TimesFM-3GoogleNew model
    Google releases TimesFM-3: 330M-parameter zero-shot multivariate forecaster

August 2026

  1. StartLux 27BStartLuxLeaderboard
    StartLux's 27B local model takes second on CAICT's MCP test, beating DeepSeek-V4-Flash
  2. GLM-5.3Zhipu AIOpen weights
    Z.ai opens GLM-5.3 weights for coding and vulnerability huntingGLM timeline →
  3. Hunyuan Hy4 previewTencentOpen weights
    Tencent open-sources Hunyuan Hy4 preview: 770B total, 49B active
  4. GLM-5.3-FlashZhipu AILeaderboard
    GLM-5.3-Flash tops OpenRouter usage, served on domestic chipsGLM timeline →
  5. Gemini Omni 1.1 FlashGoogleNew model
    Gemini Omni 1.1 Flash Adds Scene and 4K ControlsGemini timeline →
  6. WALL-SSX Square RobotNew model
    WALL-SS Simulates 60-Second Robot Tasks
  7. GLM-5.3-FlashZhipu AIOpen weights
    GLM-5.3-Flash Opens With a 1M-Token Context WindowGLM timeline →
  8. Qwen3.8-FlashAlibabaOpen weights
    Qwen3.8-Flash Opens With 6B Active ParametersQwen timeline →
  9. Gemini 3.5 TranscribeGoogleNew model
    Gemini 3.5 Transcribe Enters Public PreviewGemini timeline →
  10. Thomson-1Thomson ReutersNew model
    Thomson Reuters Reportedly Launches Qwen-Based Thomson-1
  11. Qwen3.6-27BAlibabaPlatform
    JetBrains builds local Junie on Qwen3.6-27BQwen timeline →
  12. GPT-5.6OpenAIPlatform
    GPT‑5.6 arrives in KiroOpenAI timeline →
  13. GEN-1.5Generalist AINew model
    GEN-1.5 learns tasks from seconds-long demos
  14. SHADOW 250M InstructUndisclosedOpen weights
    SHADOW 250M Fits Into a 60MB Deployment
  15. TenetHarveyNew model
    Harvey Tenet Post-Trains Kimi K3 for Legal Agents
  16. Ox AlphaUndisclosedNew model
    Anonymous Ox Alpha Offers a One-Million-Token Window
  17. Wan3.0AlibabaNew model
    Wan3.0 Generates 30-Second Video from Documents
  18. LFM2.5-DSparkLiquid AINew model
    LFM2.5-DSpark lifts inference throughput up to 3.18x via speculative decoding
  19. Qwen-UI-AgentAlibabaNew model
    Alibaba releases Qwen-UI-Agent, a foundation model for GUI agentsQwen timeline →
  20. Grok 4.6xAIPlatform
    Grok 4.6 arrives on Amazon Bedrock with a 500K context windowGrok timeline →
  21. GLM-5.3Zhipu AIAPI
    Z.ai opens GLM-5.3 API for coding and defensive securityGLM timeline →
  22. HappyShrimpAlibabaNew model
    Alibaba launches HappyShrimp AI music model
  23. Qwen3.8-27BAlibabaLeaderboard
    Third-party benchmark ranks open Qwen3.8-27B 15thQwen timeline →
  24. Qwen3.8-27BAlibabaOpen weights
    Alibaba open-sources Qwen 3.8 27B under Apache 2.0Qwen timeline →
  25. Grok 4.6xAIPlatform
    Grok 4.6 arrives in GitHub CopilotGrok timeline →
  26. GLM-5.3Zhipu AINew model
    Z.ai ships GLM-5.3 with post-training-only gainsGLM timeline →
  27. Gemini 3.7 FlashGoogleNew model
    Google ships Gemini 3.7 Flash at half price, three weeks after 3.6Gemini timeline →
  28. Palmyra X6WriterNew model
    Writer launches Palmyra X6, post-trained on Z.ai's GLM-5.2
  29. Grok 4.6xAINew model
    SpaceXAI ships Grok 4.6, tying GPT-5.6 on the AA IndexGrok timeline →
  30. OlmoEarthAi2Open weights
    AllenAI opens OlmoEarth embedding exports, with code and weights released
  31. Grok 4.6xAIPlatform
    Cognition adds Grok 4.6 to Devin's coding platformGrok timeline →
  32. Qwen3.8-2.4T-A95BAlibabaOpen weights
    Alibaba open-weights Qwen3.8-2.4T-A95B, its first Max-tier weightsQwen timeline →
  33. DeepSeek-V4-Pro-0813DeepSeekNew model
    DeepSeek V4 Pro goes live with a 1M-token tierDeepSeek timeline →
  34. SL2TGoogleNew model
    Google DeepMind ships sign-language-to-text model SL2T
  35. LFM2.5-VL-3BLiquid AINew model
    Liquid AI releases edge vision model LFM2.5-VL-3B
  36. MAI Code 1.1 FlashMicrosoftNew model
    Microsoft's MAI Code 1.1 Flash trails DeepSeek on price and performance
  37. Ishigaki-IDSONESTRUCTIONNew model
    ONESTRUCTION builds construction-BIM foundation model Ishigaki-IDS with AWS
  38. QwenAlibabaPlatform
    Alibaba's Qwen app expands: Honor Robot Phone debut and Midea integrationQwen timeline →
  39. Nemotron 3.5 LightningNVIDIAOpen weights
    Nvidia open-sources Nemotron 3.5 Lightning, a fast 30B MoE
  40. Daybreak RedOpenAIPlatform
    OpenAI's Daybreak cyber models arrive on Amazon Bedrock
  41. Daybreak BlueOpenAIPlatform
    OpenAI's Daybreak cyber models arrive on Amazon Bedrock
  42. LTX-2.5LightricksOpen weights
    LTX-2.5 ships as an open-weights video world model that runs on one desk
  43. Muse GlimmerMetaOpen weights
    Meta open-sources Muse Glimmer, a 30B model for one GPU
  44. Claude Sonnet 5AnthropicNew model
    Anthropic releases Claude Sonnet 5Claude timeline →
  45. Needle 2CactusNew model
    Cactus ships Needle 2, a 14MB on-device agentic model
  46. VoiceChat 11BNVIDIAOpen weights
    NVIDIA open-sources full-duplex speech model VoiceChat 11B
  47. Isaac 28BPokee AINew model
    Pokee AI ships Isaac 28B with a 10M-token context on a single GPU
  48. Wan 3.0AlibabaNew model
    Alibaba's Qwen app adds five free features; Wan 3.0 enters public test
  49. Cosmos 3NVIDIAOpen weights
    NVIDIA's open Cosmos 3 world models span 64B/16B/4B
  50. WeatherNextGoogleOpen weights
    DeepMind open-sources WeatherNext for cyclone forecasts
  51. Xiaomi-Robotics-1XiaomiOpen weights
    Xiaomi open-sources Xiaomi-Robotics-1, a 5B robot model
  52. Muse Spark 1.2MetaNew model
    Meta ships Muse Spark 1.2 and the Muse Code terminal coding agent
  53. Doubao 2.0ByteDancePlatform
    SAIC Roewe Jiayue 07 to run Doubao 2.0 and Momenta's R7 world model
  54. GeminiGooglePlatform
    Google Assistant retires next month; Gemini becomes Android's default assistantGemini timeline →
  55. Alpamayo 2 SuperNVIDIAOpen weights
    NVIDIA open-sources Alpamayo 2 Super: a 34B vision-language-action model for autonomous driving
  56. MAGI-2-previewSand.aiOpen weights
    Sand.ai open-sources MAGI-2-preview: a 114B/6B-active MoE video model
  57. ShieldstralMistral AIOpen weights
    Mistral releases Shieldstral, a 3B open-weight multimodal safety classifier
  58. LFM2.5-2.6BLiquid AINew model
    Liquid AI ships LFM2.5-2.6B for on-device agents
  59. DeepSeek V4-FlashDeepSeekLeaderboard
    DeepSeek V4-Flash tops OpenRouter's weekly token rankingDeepSeek timeline →
  60. Qwen3.8-MaxAlibabaNew model
    Alibaba's Qwen3.8-Max rivals top US frontier modelsQwen timeline →
  61. MiniMax H3MiniMaxOpen weights
    MiniMax open-sources H3, first open model to top a video rankingMiniMax timeline →
  62. SenseNova U1.5-LiteSenseTimeOpen weights
    SenseTime open-sources a preview of SenseNova U1.5-Lite (Chinese-language source)
  63. Instella-MoEAMDOpen weights
    AMD ships the fully open Instella-MoE: 16B total, 2.8B active
  64. DoubaoByteDancePlatform
    Tesla China adds ByteDance's Doubao — its first third-party in-car LLM
  65. Seedance 2.5ByteDancePlatform
    Seedance 2.5 video model arrives in the creation tool "Xiaoyunque"

July 2026

  1. DeepSeek V4-FlashDeepSeekNew model
    DeepSeek retunes V4-Flash to strengthen coding agentsDeepSeek timeline →
  2. Qwen-Audio-3.0-ASR-FlashAlibabaNew model
    Alibaba's Qwen upgrades speech recognition for technical termsQwen timeline →
  3. Gemini Robotics ER 2GoogleNew model
    Google DeepMind launches Gemini Robotics ER 2Gemini timeline →
  4. Grok Voice Think Fast 2.0xAINew model
    SpaceXAI ships Grok Voice Think Fast 2.0 at a 60% price premiumGrok timeline →
  5. Lyria 3.5GoogleNew model
    Google DeepMind launches Lyria 3.5 music model in Flow Music
  6. GeminiGooglePlatform
    Waymo adds Gemini and a three-screen cabin UI to Ojai robotaxisGemini timeline →
  7. MiMo-V2.5XiaomiLeaderboard
    Chinese open models top OpenRouter; Xiaomi MiMo-V2.5 is No. 1 by usage
  8. OlmoEarthAi2New model
    Ai2 launches OlmoEarth: geospatial inference at planetary scale
  9. HyWorldVLABYDNew model
    BYD's AI team debuts with the HyWorldVLA autonomous-driving world model
  10. Kimi K3Moonshot AIPlatform
    Replit rolls out model choice, starting with Kimi K3Kimi timeline →
  11. LFM2.5-EncodersLiquid AIOpen weights
    Liquid AI open-sources LFM2.5-Encoders for fast long-context inference on CPU
  12. Kimi K3Moonshot AIOpen weights
    Kimi K3 open weights land, inference providers onboard on day oneKimi timeline →
  13. MAI-Cyber-1-FlashMicrosoftNew model
    Microsoft ships its first in-house cybersecurity model, MAI-Cyber-1-Flash
  14. Ling-3.0-flashAnt GroupNew model
    Ant unveils Ling-3.0-flash, a 124B hybrid-reasoning model
  15. Claude Opus 5AnthropicNew model
    Claude Opus 5 launches, nearing Fable 5 at half the priceClaude timeline →
  16. GPT-5.6OpenAIPlatform
    GPT-5.6 Sol, Terra and Luna go GA on Amazon BedrockOpenAI timeline →
  17. UniWorld-ViewRabbitpreLeaderboard
    China's UniWorld-View world model tops a Fei-Fei Li team benchmark
  18. Fugu-Ultra v1.1Sakana AINew model
    Sakana AI ships Fugu-Ultra v1.1, up to +7.9 on benchmarks
  19. Llama 3.1 405BMetaPlatform
    Petals lists 405B Llama support for volunteer-GPU inference
  20. FabriVLAYouibotLeaderboard
    Youibot's FabriVLA tops Meta-World MT50 with 1B parameters, beating π0
  21. AntaresCiscoOpen weights
    Cisco open-sources Antares security models for vulnerability localization
  22. Qwen-Image-3.0AlibabaNew model
    Alibaba releases Qwen-Image-3.0 with 4.5× longer text input (Chinese-language source)Qwen timeline →
  23. COOWAMCowa RobotNew model
    Coowa shows a dual-layer agent world model, COOWAM, at WAIC
  24. Gemini 3.6 FlashGoogleNew model
    Google releases Gemini 3.6 Flash, 3.5 Flash-Lite and 3.5 Flash CyberGemini timeline →
  25. Gemini 3.5 Flash-LiteGoogleNew model
    Google releases Gemini 3.6 Flash, 3.5 Flash-Lite and 3.5 Flash CyberGemini timeline →
  26. Gemini 3.5 Flash CyberGoogleNew model
    Google releases Gemini 3.6 Flash, 3.5 Flash-Lite and 3.5 Flash CyberGemini timeline →
  27. MiniCPM-RobotModelBestOpen weights
    OpenBMB Open-Sources 1.5B General-Purpose VLA Model MiniCPM-Robot
  28. Qwen3.8-MaxAlibabaNew model
    Alibaba Previews 2.4-Trillion-Parameter Qwen3.8Qwen timeline →
  29. SenseNova U1 ProSenseTimeNew model
    SenseNova U1 Pro enters preview with up to 8K output
  30. JoyAI-TalkerJD.comNew model
    JD launches JoyAI-Talker and JoyAI-Video-Edit
  31. JoyAI-Video-EditJD.comNew model
    JD launches JoyAI-Talker and JoyAI-Video-Edit
  32. Kimi K3Moonshot AILeaderboard
    Kimi K3 tops Text Arena science query rankings; community praises cost-performanceKimi timeline →
  33. SenseNova U1 ProSenseTimeNew model
    SenseTime unveils SenseNova U1 Pro, a multimodal flagship targeting long-horizon tasks
  34. Tianpuyue V4.7QuwanNew model
    Tianpuyue music model V4.7 released, boosting remix and cover capabilities
  35. Kimi K3Moonshot AILeaderboard
    Kimi K3 wins Musk’s praise, beats Sonnet 5 on Simple Bench, tops community evalsKimi timeline →
  36. Kimi K3Moonshot AINew model
    Kimi K3: full specs published, 2.8T parameters, weights by July 27Kimi timeline →
  37. Kimi K3Moonshot AILeaderboard
    ArtificialAnalysis: Kimi K3 scores 57, ranking 4th of 189 modelsKimi timeline →
  38. Kimi K3Moonshot AILeaderboard
    Arena WebDev board: Kimi K3 takes first at 1679, 48 points clear of Claude Fable 5Kimi timeline →
  39. Kimi K3Moonshot AINew model
    Kimi K3 ships on web and app — the API still runs K2.7Kimi timeline →
  40. DW0.5DexmalNew model
    Force Lingji Releases DW0.5: A World Model That Coaches VLAs
  41. Lumo-2AstribotNew model
    Stardust Intelligence Releases the Embodied Foundation Model Lumo-2
  42. Soofi SGerman research consortiumOpen weights
    German Consortium Open-Sources Soofi S: 31.6B Total, Only 3.2B Active
  43. MiniCPMModelBestPlatform
    ModelBest's MiniCPM to Ship on Samsung Phones
  44. InklingThinking Machines LabOpen weights
    Thinking Machines Lab Releases the Multimodal Open Model Inkling
  45. Doubao-Seed-EvolvingByteDanceNew model
    ByteDance Seed Upgrades Its Model: 1M Context and a Cost Promotion
  46. GPT-RedOpenAINew model
    OpenAI Releases GPT-Red, an Internal Automated Red-Teaming Model
  47. Qwen-Audio-3.0-RealtimeAlibabaNew model
    Qwen Releases the Real-Time Voice Model Qwen-Audio-3.0-RealtimeQwen timeline →
  48. Xiaomi-Robotics-U0XiaomiOpen weights
    Xiaomi Open-Sources the Embodied Generation Model Xiaomi-Robotics-U0