AI Trend Notifier
EN한

$ ls wiki/models/ --sort=date

모델 릴리스 타임라인

Every model this wiki tracks, newest first. Each one links to its page, where the claim is cited to the announcement it came from.

121 models · 2026-04-07 → 2026-10-03 · 2 without a recorded date

October 20261

  1. 03Kolibri-1releasedaleph-alpha's 2026-10-03 open-weight release, and the firstAleph Alpha

September 202634

  1. 30Gemini 4released"Significantly larger than any prior Gemini model"Google DeepMind
  2. 30Gemini 4 Argonreleasedgoogle-deepmind's 2026-09-30 frontier release, announced by SVPGoogle DeepMind
  3. 29GPT-6.1 SolreleasedOpenAI's 2026-09-29 DevDay release, seven days afterOpenAI
  4. 29NVIDIA Kumo TabularreleasedAn open foundation model for tabular classification and regression, publishedNVIDIA
  5. 28Claude Sonnet 5.5releasedAnthropic's 2026-09-28 Sonnet release, six days afterAnthropic
  6. 23Gemini 3.8 Flash TTSreleasedgoogle-deepmind's expressive text-to-speech model, announcedGoogle DeepMind
  7. 23Gemini 3.8 Flash-Lite TTSreleasedgoogle-deepmind's high-volume text-to-speech model, announcedGoogle DeepMind
  8. 23Muse Realtime AvatarannouncedStated as embodiment for a conversational agent: it pairs with Muse RealtimeMeta AI
  9. 22Claude Opus 5.5releasedAnthropic's 2026-09-22 release of the Opus line, and the first model on thisAnthropic
  10. 22GPT-6 LunareleasedOpenAI's 2026-09-22 fast, cost-efficient model of the GPT-6 series, releasedOpenAI
  11. 22GPT-6 SolreleasedOpenAI's 2026-09-22 cost-efficient high-end model of the GPT-6 series,OpenAI
  12. 22MiMo-V2.6-Proreleasedxiaomi's 2026-09-22 flagship: a 1.02T-parameter sparseXiaomi
  13. 22Ming-Image-0.1-DesignreleasedThe stated target is text-rich visual design rather than general imageAnt Group
  14. 21Grok 4.7releasedxai's flagship coding and knowledge-work model, releasedxAI
  15. 18Grok Voice Transcribe 2.0releaseddetection and mid-recording language switching handled in a single pass;xAI
  16. 17Ternary Bonsai 2 27BreleasedLocal and on-device inference is the entire proposition: 5.93 GB under Apache 2.0, in GGUF and MLX, which is NVIDIA and Apple loc…Prismml
  17. 15Gemini 3.8 LivereleasedPositioned as the scale-and-cost-efficiency option of the pair, built forGoogle DeepMind
  18. 15Gemini 3.8 Live Extended ThinkingreleasedPresented as undercutting both while topping them on quality — 27% below theGoogle DeepMind
  19. 15JevreleasedJev does not generate text. The caller sends structured or natural-languageTypesafe
  20. 11Fugu MaxreleasedFugu Max is not a single trained network.Sakana AI
  21. 11Fugu Ultra v2releasedThe capability tier of the same orchestration architecture as fugu-maxSakana AI
  22. 10DeepSeek V4.1-FlashreleasedDeepSeek describes the model as the first in a Causal Encoder-Decoder (CED)DeepSeek
  23. 08ChatGPT Images 2.5releasedImage generation and editing inside ChatGPT, with two new interaction surfacesOpenAI
  24. 08GPT-Image-2.5 FlarereleasedOpenAI positions this as the default choice for most applications.OpenAI
  25. 08GPT-Image-2.5 SunburstreleasedPremium visual workflows that benefit from tighter control across edits —OpenAI
  26. 07Qwen-Drive-1.0-4BreleasedRetains the pretrained Qwen3.5-4B vision-language architecture and adds a BEVAlibaba Qwen
  27. 03AstrareleasedOpenAI's next major model, named publicly for the first time on 2026-08-01 in aOpenAI
  28. 03K2 HorizonreleasedA family of six models from the Institute of Foundation Models, releasedIfm
  29. 03WeatherNext 3releasedOperational forecasting shipped directly into consumer surfaces — Search, theGoogle DeepMind
  30. 02Gemini 3.8 FlashreleasedPositioned as a workhorse, the same word Google used for 3.7 Flash, with theGoogle DeepMind
  31. 02Gemini 3.8 Flash CyberreleasedAutonomous vulnerability discovery at agentic scaleGoogle DeepMind
  32. 02Muse Spark 1.3releasedTwo published reasoning variants exist: max (top reasoning) and xhighMeta AI
  33. 01Claude Fable 5.1releasedAnthropic's 2026-09-01 point release of the Fable line, shipped together withAnthropic
  34. 01Muse Voice TranscribereleasedReal-time dictation and live transcription, where the three jobs Meta collapsesMeta AI

August 202622

  1. 28Hy4 previewreleasedHy4 preview enters at #10 with 5.23% ±1.31%, listed under TencentTencent
  2. 27Gemini Omni 1.1 FlashreleasedThe release is framed around control rather than generation quality, and eachGoogle DeepMind
  3. 26Gemini 3.5 TranscribereleasedDictation and note capture is the deployment: Gboard Rambler ships on it, andGoogle DeepMind
  4. 26GLM-5.3-FlashreleasedCoding and agentic work is where every published figure sits (DeepSWE,Z.ai
  5. 25Granite 4.2releasedLocal and on-premise enterprise deployment is the stated framing: Apache 2.0IBM
  6. 25Qwen3.8-Flash-NextreleasedAgentic coding and tool use, on the evidence of what was benchmarked: everyAlibaba Qwen
  7. 21DeepSeek V4-Flash-Vision-ExpreleasedAn experimental multimodal (vision) build of DeepSeek's flagship text-onlyDeepSeek
  8. 14GLM-5.3releasedCoding and long-horizon agentic work, delivered through a subscription codingZ.ai
  9. 13Gemini 3.7 Flashreleasedbenchmark selection — enterprise workflow automation, coding, long-contextGoogle DeepMind
  10. 13MiniMax Music 3.0releasedComplete-song generation from a lyric sheet and a description, in one pass ratherMiniMax
  11. 12SL2TreleasedTwo models, split across the device boundaryGoogle DeepMind
  12. 11Nemotron 3.5 LightningreleasedNVIDIA's stated framing is explicitly subordinate: a frontier reasoning modelNVIDIA
  13. 10GPT-5.6-CyberreleasedAuthorized vulnerability research, exploit validation and security testing,OpenAI
  14. 10Muse GlimmerreleasedLocal coding agents and other always-on agent workflows — the stated designMeta AI
  15. 07Grok Imagine Image 2.0releasedEditing is the release, and the feature list is an editing tool rather than axAI
  16. 06Ling-3.0-tinyreleasedThe stated target is local and resource-constrained agent deployment: aAnt Group
  17. 06WeatherNext CyclonesreleasedOperational tropical-cyclone forecasting — track, intensity and size, worldwide.Google DeepMind
  18. 05Muse Spark 1.2releasedMuse Code — terminal coding agent, beta on macOS and Linux.Meta AI
  19. 04LFM2.5-2.6BreleasedLiquid AI's stated target is agents that run where the data is: the model "plans,Liquid AI
  20. 04Shieldstral 1.0releasedDrop-in guardrail in front of or behind a generating model, for text, images orMistral AI
  21. 03Qwen 3.8 27BreleasedAlibaba's stated positioning is on-premise deployment: this is described as theAlibaba Qwen
  22. 03Qwen-Image-2.1releasedThe README states text-to-image at native 2K, image editing with up toAlibaba Qwen

July 202627

  1. 31DeepSeek V4-FlashreleasedArtificial Analysis scored the model on its own index.DeepSeek
  2. 31GWM Worlds 2announcedThe stated interface is WorldPrompt, described in two passes as a promptingRunway
  3. 31MiniMax H3releasedThe release is partial by MiniMax's own descriptionMiniMax
  4. 30Gemini Robotics ER 2releasedTask orchestration — planning and sequencing over a VLA acting layerGoogle DeepMind
  5. 29Grok Voice Think Fast 2.0releasedSpeech-to-speech voice agents. The stated design choice is that Think Fast 2.0xAI
  6. 29Lyria 3.5releasedMusic creation inside Google Flow Music.Google DeepMind
  7. 28DeepSeek V4-Pro-0813releasedThis page published a retirement that is not happening, and the correction isDeepSeek
  8. 28Gemini Robotics 2releasedHumanoid whole-body tasks requiring locomotion and manipulation in one policy: walking,Google DeepMind
  9. 25Grok 4.6releasedGrok 4.6 is absent from the 2026-08-09 Artificial Analysis snapshot andxAI
  10. 22Cosmos-H-DreamsreleasedCosmos-H-Dreams is an action-conditioned world foundation model for surgical robotics simulation.NVIDIA
  11. 21Gemini 3.5 Flash CyberreleasedLarge-scale automated vulnerability scanningGoogle DeepMind
  12. 21Gemini 3.5 Flash-LitereleasedHigh-throughput, low-latency agentic pipelines (agentic search, document processing)Google DeepMind
  13. 21Gemini 3.6 FlashreleasedCoding and software engineering (SWE-Bench Pro 58.7%)Google DeepMind
  14. 21Laguna S 2.1releasedAgentic coding — long-horizon software tasks, which is what the benchmark set PoolsidePoolside
  15. 19Qwen 3.8 MaxreleasedThe figure that makes this release readable as an engineering event rather than aAlibaba Qwen
  16. 17Kimi K2.8 PreviewreleasedCoding and agentic work, which is what both surfaces it shipped to are for —Moonshot AI
  17. 16Kimi K3releasedLong-horizon agentic coding (K3 Swarm Max variant)Moonshot AI
  18. 15InklingreleasedThe lab's own stated role for the model is a balanced, multimodal, customizableThinking Machines
  19. 10AlphaEvolvereleasedDesigning advanced algorithms via evolutionary searchGoogle DeepMind
  20. 08Claude Opus 5releasedLMArena's weekly snapshot puts the top three places with Anthropic, and theAnthropic
  21. 08GPT-Live-1releasedConversational AI / personal assistant voice interactionOpenAI
  22. 08Grok 4.5releasedxAI's enterprise deployment of the V9-class model.xAI
  23. 08Robostral NavigatereleasedRobostral Navigate predicts robot movement via pointing: given the current camera frame, it outputs the pixel coordinates of the …Mistral AI
  24. 07Muse ImagereleasedImage editing (single reference)Meta AI
  25. 07Muse VideoreleasedBuilt on the same pretraining base as muse-image (shared foundation)Meta AI
  26. 04Mistral Large 3releasedCEO Arthur Mensch confirms the model as "fat but sparse" MoE; ARR $400M+Mistral AI
  27. 01Leanstral 1.5releasedReal-world code verification: found 5 previously unknown bugs across 57 open-source repositoriesMistral AI

June 202616

  1. 30Claude SciencereleasedClaude Science is a scientific research workbench that gives researchers a unified environment for computational work.Anthropic
  2. 30Claude Sonnet 5releasedDrop-in upgrade for agentic coding workflows running on Sonnet 4.6 — same pricing tier, significantly higher capabilityAnthropic
  3. 30LongCat-2.0releasedLongCat-2.0 had been running quietly on OpenRouter under the codename "Owl Alpha" before its identity was revealed at open-source…Meituan
  4. 30Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image)releasedHigh-throughput API image generation (cheapest Google option per image)Google DeepMind
  5. 16Grok V9-MediumreleasedxAI's coding-focused foundation model — 1.5-trillion parameters, ~3× larger than the prior production Grok.xAI
  6. 13GLM-5.2releasedGLM-5.2 introduces IndexShare — an architectural optimization to sparse attention that reduces per-token compute by 2.9× at the f…Z.ai
  7. 11DiffusionGemmareleasedGoogle DeepMind's first open-weight text diffusion model, built on the Gemma 4 26B MoE backbone.Google DeepMind
  8. 09Claude Fable 5releasedThe first third-party measurement of what this model actually sells.Anthropic
  9. 03Gemma 4 12BreleasedNo cloud compute required for inferenceGoogle DeepMind
  10. 03Grok Imagine Video 1.5 (Preview)releasedCamera moves, atmospheric effects, physics-consistent animationxAI
  11. 02GPT-5.6 Sol (and Terra, Luna)releasedA new service tier, not a new model: GPT-5.6 Sol at up to 14× the speed ofOpenAI
  12. 02MAI-Code-1 / MAI-Code-1-FlashreleasedA coding-specialized model that Microsoft trained inside GitHub Copilot's actual production harness, optimized directly for Copil…Microsoft
  13. 02MAI-Thinking-1releasedA reasoning-specialized model developed in-house by Microsoft's MAI (Microsoft AI) team.Microsoft
  14. 01Cosmos 3 SuperreleasedThe world's first fully open omnimodal physical AI foundation model.NVIDIA
  15. 01MiniMax M3releasedMSA (MiniMax Sparse Attention): proprietary sparse attention mechanismMiniMax
  16. 01Project PolarisannouncedMicrosoft's first in-house frontier-class coding AI model.Microsoft

May 202611

  1. 28Claude Opus 4.8releasedUncritically reporting flawed results: 0% (first Claude model to achieve zero)Anthropic
  2. 21Co-Scientist (Google DeepMind)releasedScientific hypothesis generation: Given a research question or grant proposal summary, generates and ranks novel hypothesesGoogle DeepMind
  3. 19Gemini 3.5 FlashreleasedMulti-step agent workflows (MCP Atlas 83.6%)Google DeepMind
  4. 19Gemini 3.5 ProreleasedAnnounced at Google I/O, June GA promisedGoogle DeepMind
  5. 19Gemini OmniannouncedGemini Omni represents "a leap forward in world understanding, multimodality and editing" — Google's framing at I/O.Google DeepMind
  6. 19Gemini SparkannouncedGemini Spark is a 24/7 cloud-based personal agent that takes actions on behalf of users even when they're offline.Google DeepMind
  7. 14Grok BuildreleasedA long-running autonomous execution mode shipped June 22, 2026.xAI
  8. 12Gemma 3nreleasedKey engineering contribution of Gemma 3n: Per-Layer Embeddings (PLE) dramatically reduce RAM footprint compared to traditional em…Google DeepMind
  9. 07GPT-5.5 InstantreleasedA claim of simultaneous improvement across three axes (intelligence, clarity, personalization).OpenAI
  10. 07GPT-Realtime-2 (OpenAI)releasedThree models released together:OpenAI
  11. 06Grok 4.1 Fast (xAI)announcedNew managed tool infrastructure requiring minimal developer setup:xAI

April 20268

  1. 29Mistral Medium 3.5releasedPending further detail from the full announcement materials.Mistral AI
  2. 24DeepSeek V4releasedV4-Pro's GA build now has its own page — deepseek-v4-pro-0813.DeepSeek
  3. 22Deep Research MaxreleasedNightly due diligence reports: Asynchronous background run, exhaustive synthesis by morningGoogle DeepMind
  4. 16Claude Opus 4.7releasedThoroughness / consistencyAnthropic
  5. 16GPT-RosalindreleasedStronger reasoning about molecules, proteins, genes, and disease biology vs.OpenAI
  6. 15Gemini Robotics ER 1.6releasedIndustrial inspection (energy plants, manufacturing, chemical facilities)Google DeepMind
  7. 08Muse Spark (1.0 / 1.1)releasedMajor upgrade released alongside the Meta Model API — Meta's first paid external AI product.Meta AI
  8. 07Claude Mythos PreviewreleasedNot applicable for general users.Anthropic

No date recorded

119 of 121 model pages state a release or announcement date. These do not, so they are listed here rather than placed at a guessed position.