Codestral
Mistral AI's first code model, a 22B open-weight system trained on more than 80 programming languages.
Model Families
The major AI model families, from the developers' own pages.
90 entries, all primary-sourced
Kling (Kuaishou video-generation model)
Kuaishou's text-to-video model, unveiled in June 2024, generating up to two-minute clips at 1080p and 30 fps.
Runway Gen-3 Alpha
Runway's 2024 video model that improved fidelity and motion over Gen-2 and was framed as a step toward general world models.
Oasis (Decart and Etched playable AI world model)
A real-time, playable AI-generated Minecraft-like world, produced frame by frame by a neural model with no game engine.
Evo (DNA foundation models)
Arc Institute's family of genomic foundation models that read and generate DNA, RNA, and protein sequences from raw nucleotides.
Claude Opus 4.8
Anthropic's May 2026 flagship, an Opus 4.7 upgrade with sharper agentic judgment, faster fast mode, and dynamic workflows.
NVIDIA Cosmos 3
NVIDIA's open foundation model for physical AI that unifies vision reasoning, world generation, and action across robots and vehicles.
Anthropic releases Claude Fable 5 and Mythos 5
Anthropic launched Claude Fable 5, a frontier Mythos-class model, plus a safeguard-lifted Mythos 5 for authorized partners.
Google DeepMind launches Gemini 3.5 Live Translate
Google DeepMind shipped Gemini 3.5 Live Translate, a streaming speech-to-speech model that translates audio while the speaker is still talking.
Z.ai releases GLM-5.2 open-weight coding model
Z.ai released GLM-5.2, an MIT-licensed open-weight Mixture-of-Experts model with a 1M-token context aimed at agentic coding.
Mistral releases OCR 4 document-intelligence model
Mistral launched OCR 4, a document-intelligence model with bounding boxes, block classification, confidence scores, and 170-language support.
OpenAI previews GPT-5.6 Sol, Terra, and Luna under US government coordination
OpenAI's June 26, 2026 GPT-5.6 preview spans Sol, Terra, and Luna, initially limited to trusted partners at the US government's request.
Anthropic releases Claude Sonnet 5
Anthropic's June 30, 2026 midsize model nears Opus 4.8 on agentic tasks at a fraction of the cost, default for Free and Pro users.
Google ships Nano Banana 2 Lite image model and Gemini Omni Flash video model
Google's June 30, 2026 launch pairs a 4-second, low-cost image model with a natively multimodal video generation model in public preview.
GPT-Live: OpenAI's full-duplex voice models power a new ChatGPT Voice
OpenAI launches GPT-Live, full-duplex voice models that listen and speak at once and delegate hard questions to a frontier model in the background.
GPT-5.6: OpenAI's Sol, Terra, and Luna family goes GA
OpenAI releases the GPT-5.6 family - flagship Sol, balanced Terra, and low-cost Luna - claiming new highs in agentic work at lower cost.
Grok 4.5: SpaceXAI's coding-focused flagship trained alongside Cursor
SpaceXAI launches Grok 4.5, a coding and agentic flagship trained with Cursor, priced at $2/$6 per million tokens and served at 80 TPS.
Kimi K3: Moonshot AI's 2.8T-parameter open frontier model
Moonshot AI introduces Kimi K3, a 2.8 trillion parameter model with 1M-token context, calling it the world's first open 3T-class model.
Google launches Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber
Google shipped three Gemini models on July 21, 2026, led by 3.6 Flash at $1.50 in and $7.50 out per million tokens.
Poolside releases Laguna S 2.1 open-weight coding model
Poolside released Laguna S 2.1 on July 21, 2026, a 118B mixture-of-experts coding model with open weights on Hugging Face.
Black Forest Labs announces FLUX 3 multimodal model
Black Forest Labs announced FLUX 3 on July 23, 2026, one model trained jointly on images, video, and audio plus robot actions.
Anthropic releases Claude Opus 5
Anthropic released Claude Opus 5 on July 24, 2026, holding Opus 4.8 pricing while claiming Fable 5 intelligence at half the cost per task.
DeepSeek ships V4-Flash-0731, a retrained agentic build
DeepSeek released the MIT-licensed DeepSeek-V4-Flash-0731 on July 31, 2026, beating its own V4-Pro preview on every published benchmark.
OpenAI shipped August updates to GPT-5.6 Sol and Luna and changed the ChatGPT default model
OpenAI updated GPT-5.6 Sol and Luna, reporting about 60 percent fewer factuality errors and a new ChatGPT default.
Meta released Muse Glimmer, a 30B open-weight agentic model built to run on consumer hardware
Meta released Muse Glimmer, a 30 billion parameter multimodal agentic model under Apache 2.0 that runs on a single GPU.
NVIDIA releases Nemotron 3.5 Lightning and the NeMo Switchyard model router
NVIDIA shipped Nemotron 3.5 Lightning, a 30B open MoE for high-volume agent tasks, plus an open-source model router.
Alibaba published open weights for Qwen3.8-2.4T-A95B, its first Max-class open model
Alibaba published open weights for Qwen3.8-2.4T-A95B, the first Qwen-Max-class model released for download.
SpaceXAI releases Grok 4.6 for long-running agents and interactive work
SpaceXAI released Grok 4.6, an agent-focused update to Grok 4.5 priced at 2 dollars per million input tokens.
DeepSeek moved V4-Pro to general availability and split its API into peak and off-peak pricing
DeepSeek took V4-Pro to general availability with tiered reasoning effort and split API pricing into peak and off-peak rates.
Google released Gemini 3.7 Flash three weeks after Gemini 3.6 Flash
Google shipped Gemini 3.7 Flash three weeks after 3.6 Flash, with large coding and agent gains at introductory pricing.
Z.ai releases GLM-5.3, a post-trained GLM-5.2 with emergent vulnerability-discovery ability
Z.ai shipped GLM-5.3, built by post-training GLM-5.2 alone, and says it has already found 2,436 real vulnerabilities.
Alibaba released Qwen3.8-Flash-Next open weights
Alibaba open-weighted Qwen3.8-Flash-Next, a 125B mixture-of-experts model previewing the Qwen4 architecture.
Anthropic released Claude Fable 5.1 and Claude Mythos 5.1
Anthropic shipped Claude Fable 5.1 and Mythos 5.1, cutting cache read prices 75 percent to 0.25 dollars per million tokens.
Google released Gemini 3.8 Flash and Gemini 3.8 Flash Cyber
Google released Gemini 3.8 Flash and a restricted Flash Cyber variant, its third Flash model in roughly six weeks.
Meta released Muse Spark 1.3
Meta released Muse Spark 1.3, an agentic coding model using about 20 percent fewer tool calls than Muse Spark 1.2.
OpenAI released GPT-6 Astra
OpenAI launched GPT-6 Astra, reporting 98 percent on FrontierMath Tier 4 and a perfect 100 percent on ExploitBench.
OpenAI released GPT Image 2.5 in two variants, Sunburst and Flare
OpenAI released GPT Image 2.5 Sunburst and Flare in its API on Sept 8, 2026, adding xhigh and max quality settings for image work.
DeepSeek shipped V4.1-Flash, a 552B-parameter encoder-decoder MoE with native vision
DeepSeek released V4.1-Flash on Sept 10, 2026, a 552B MoE model that beats V4-Pro while activating only 8B/16B parameters.
Sakana AI released Fugu Ultra v2 and Fugu Max, models that orchestrate other models
Sakana AI shipped Fugu Ultra v2 and Fugu Max on Sept 11, 2026, orchestrator models that route tasks across a pool of other models.
Google released Gemini 3.8 Live and 3.8 Live Extended Thinking voice models
Google released Gemini 3.8 Live speech-to-speech models, one of which reasons while it talks, plus Gemini 3.5 Transcribe.
Alibaba released Qwen3.8-Omni-Flash, a 1M-token omni-modal model
Alibaba added Qwen3.8-Omni-Flash to Model Studio, taking text, image, audio and video input across a 1M-token context window.
PrismML released Ternary Bonsai 2 27B, a 5.9GB ternary-weight Qwen3.8 model
PrismML released Ternary Bonsai 2 27B under Apache 2.0, compressing Qwen3.8 27B to 5.9GB with 98.2 percent of its benchmark score.