AI Model Releases Timeline: Latest Launches, Updated Daily
AI Model Releases 2026 Timeline
A chronological timeline of major AI model releases across 2025 and 2026 — LLMs, reasoning models, open-weights families, and image & video generators, each with its date, why it mattered, and the official announcement.
Last updated · refreshed automatically from the OpenRouter models API
Newest model launches — live
Every new model on the OpenRouter API (GPT, Claude, Gemini, Grok, and 300+ more) appears here automatically with its exact launch date, plus trending open-weights releases from Hugging Face.
Compiled automatically from the OpenRouter models API · last refreshed November 22, 2026.
-
Nov 20, 2026 · Mistral
Mistral: Pixtral Large 2502
API launchNew on the Mistral API — 128,000-token context — $2.00 in / $6.00 out per million tokens.
-
Nov 18, 2026 · Google
Google: Gemini 4 Neon
API launchNew on the Google API — 1,048,576-token context — $3.00 in / $15.00 out per million tokens.
-
Nov 17, 2026 · OpenAI
OpenAI: GPT-6.2 Astra Pro
API launchNew on the OpenAI API — 1,050,000-token context — $8.00 in / $40.00 out per million tokens.
-
Nov 17, 2026 · OpenAI
OpenAI: GPT-6.2 Astra
API launchNew on the OpenAI API — 1,050,000-token context — $8.00 in / $40.00 out per million tokens.
-
Nov 17, 2026 · OpenAI
OpenAI: GPT-6.2 Sol Pro
API launchNew on the OpenAI API — 1,050,000-token context — $2.00 in / $10.00 out per million tokens.
-
Nov 17, 2026 · OpenAI
OpenAI: GPT-6.2 Sol
API launchNew on the OpenAI API — 1,050,000-token context — $2.00 in / $10.00 out per million tokens.
-
Nov 16, 2026 · Tencent
Tencent: Hy4 flash
API launchNew on the Tencent API — 1,048,576-token context — $0.04 in / $0.14 out per million tokens.
-
Nov 16, 2026 · Qwen
Qwen: Qwen4 Max
API launchNew on the Qwen API — 1,000,000-token context — $1.42 in / $4.28 out per million tokens.
API reference — every 2026 frontier model, ready to paste
Model IDs, context windows and per-million token prices pulled from the OpenRouter models API on November 22, 2026. "API since" is first public API availability, which is often earlier than the marketing announcement.
| Model | Model ID | Context | $/M in | $/M out | API since |
|---|---|---|---|---|---|
| GPT-6.2 Astra Pro | openai/gpt-6.2-astra-pro | 1,050,000 | $8.00 | $40.00 | Nov 17, 2026 |
| GPT-6.2 Astra | openai/gpt-6.2-astra | 1,050,000 | $8.00 | $40.00 | Nov 17, 2026 |
| GPT-6.2 Sol Pro | openai/gpt-6.2-sol-pro | 1,050,000 | $2.00 | $10.00 | Nov 17, 2026 |
| GPT-6.2 Sol | openai/gpt-6.2-sol | 1,050,000 | $2.00 | $10.00 | Nov 17, 2026 |
| GPT-6.1 Terra Pro | openai/gpt-6.1-terra-pro | 1,050,000 | $1.00 | $5.00 | Oct 26, 2026 |
| Claude Fable 5.2 | anthropic/claude-fable-5.2 | 1,000,000 | $8.00 | $40.00 | Nov 9, 2026 |
| Claude Haiku 5.5 | anthropic/claude-haiku-5.5 | 1,000,000 | $0.10 | $0.50 | Oct 7, 2026 |
| Claude Sonnet 5.5 | anthropic/claude-sonnet-5.5 | 1,000,000 | $2.00 | $10.00 | Sep 28, 2026 |
| Claude Opus 5.5 | anthropic/claude-opus-5.5 | 1,000,000 | $4.00 | $20.00 | Sep 22, 2026 |
| Claude Fable 5.1 | anthropic/claude-fable-5.1 | 1,000,000 | $10.00 | $50.00 | Sep 1, 2026 |
| Gemini 4 Neon | google/gemini-4-neon | 1,048,576 | $3.00 | $15.00 | Nov 18, 2026 |
| Gemini 4 Hydrogen | google/gemini-4-hydrogen | 1,048,576 | $1.50 | $7.50 | Nov 12, 2026 |
| Gemini 4 Carbon | google/gemini-4-carbon | 1,048,576 | $2.00 | $10.00 | Nov 10, 2026 |
| Gemini 4 Argon | google/gemini-4-argon | 1,048,576 | $2.00 | $10.00 | Oct 13, 2026 |
| Gemini 3.8 Flash | google/gemini-3.8-flash | 1,048,576 | $0.82 | $4.12 | Sep 2, 2026 |
| Grok 4.7 | x-ai/grok-4.7 | 500,000 | $2.00 | $6.00 | Sep 21, 2026 |
| Grok 4.6 | x-ai/grok-4.6 | 500,000 | $2.00 | $6.00 | Aug 12, 2026 |
| DeepSeek V4.1 Flash 1017 | deepseek/deepseek-v4.1-flash-1017 | 1,048,576 | $0.15 | $0.60 | Oct 17, 2026 |
| DeepSeek V4.1 Flash | deepseek/deepseek-v4.1-flash | 1,048,576 | $0.05 | $0.20 | Sep 10, 2026 |
| DeepSeek V4 Flash Vision Exp | deepseek/deepseek-v4-flash-vision-exp | 1,048,576 | $0.22 | $0.65 | Aug 21, 2026 |
| DeepSeek V4 Pro 0813 | deepseek/deepseek-v4-pro-0813 | 1,048,576 | $1.32 | $3.96 | Aug 12, 2026 |
| DeepSeek V4 Flash 0731 | deepseek/deepseek-v4-flash-0731 | 1,048,576 | $0.04 | $0.30 | Jul 31, 2026 |
| Qwen4 Max | qwen/qwen4-max | 1,000,000 | $1.42 | $4.28 | Nov 16, 2026 |
| Qwen3.8 Max Prime | qwen/qwen3.8-max-prime | 1,000,000 | $4.00 | $12.00 | Sep 23, 2026 |
| Qwen3.8 Omni Flash | qwen/qwen3.8-omni-flash | 1,000,000 | $0.15 | $0.47 | Sep 21, 2026 |
| Qwen3.8 Max (0902) | qwen/qwen3.8-max-0902 | 1,000,000 | $2.00 | $6.00 | Sep 3, 2026 |
| Qwen3.8 Flash | qwen/qwen3.8-flash | 1,000,000 | $0.15 | $0.47 | Aug 26, 2026 |
| Kimi K3.1 Pro | moonshotai/kimi-k3.1-pro | 1,048,576 | $0.25 | $8.80 | Oct 23, 2026 |
| GLM 5.3 FlashX Legal | z-ai/glm-5.3-flashx-legal | 1,048,576 | $0.37 | $1.25 | Nov 2, 2026 |
| GLM 5.3 Prime | z-ai/glm-5.3-prime | 1,000,000 | $1.68 | $5.28 | Sep 23, 2026 |
| GLM 5.3 FlashX | z-ai/glm-5.3-flashx | 1,048,576 | $0.37 | $1.25 | Sep 18, 2026 |
| GLM 5.3 Flash | z-ai/glm-5.3-flash | 1,048,576 | $0.04 | $0.08 | Aug 26, 2026 |
| GLM 5.3 | z-ai/glm-5.3 | 1,048,576 | $0.07 | $1.40 | Aug 18, 2026 |
| Pixtral Large 2502 | mistralai/pixtral-large-2502 | 128,000 | $2.00 | $6.00 | Nov 20, 2026 |
| Mistral Large 4 | mistralai/mistral-large-4-0 | 1,048,576 | $0.70 | $2.10 | Oct 6, 2026 |
Machine-readable: this page's release feed is available as JSON at
/tools/ai-releases — date, name,
kind, summary and link per release, refreshed from provider APIs.
Curated timeline of major releases
Hand-curated sections, plus API launches compiled automaticallyNovember 2026 — latest AI model releases
Newest first. Dates are first public API availability, compiled automatically from the OpenRouter models API.
-
Nov 20, 2026 · Mistral
Mistral: Pixtral Large 2502
API launchNew on the Mistral API — 128,000-token context — $2.00 in / $6.00 out per million tokens.
-
Nov 18, 2026 · Google
Google: Gemini 4 Neon
API launchNew on the Google API — 1,048,576-token context — $3.00 in / $15.00 out per million tokens.
-
Nov 17, 2026 · OpenAI
OpenAI: GPT-6.2 Astra Pro
API launchNew on the OpenAI API — 1,050,000-token context — $8.00 in / $40.00 out per million tokens.
-
Nov 17, 2026 · OpenAI
OpenAI: GPT-6.2 Astra
API launchNew on the OpenAI API — 1,050,000-token context — $8.00 in / $40.00 out per million tokens.
-
Nov 17, 2026 · OpenAI
OpenAI: GPT-6.2 Sol Pro
API launchNew on the OpenAI API — 1,050,000-token context — $2.00 in / $10.00 out per million tokens.
-
Nov 17, 2026 · OpenAI
OpenAI: GPT-6.2 Sol
API launchNew on the OpenAI API — 1,050,000-token context — $2.00 in / $10.00 out per million tokens.
-
Nov 16, 2026 · Tencent
Tencent: Hy4 flash
API launchNew on the Tencent API — 1,048,576-token context — $0.04 in / $0.14 out per million tokens.
-
Nov 16, 2026 · Qwen
Qwen: Qwen4 Max
API launchNew on the Qwen API — 1,000,000-token context — $1.42 in / $4.28 out per million tokens.
-
Nov 12, 2026 · Google
Google: Gemini 4 Hydrogen
API launchNew on the Google API — 1,048,576-token context — $1.50 in / $7.50 out per million tokens.
-
Nov 10, 2026 · Baidu
Baidu: ERNIE-Light-DT-1.0
API launchNew on the Baidu API — 1,024,000-token context — $1.13 in / $3.94 out per million tokens.
-
Nov 10, 2026 · Google
Google: Gemini 4 Carbon
API launchNew on the Google API — 1,048,576-token context — $2.00 in / $10.00 out per million tokens.
-
Nov 9, 2026 · Anthropic
Anthropic: Claude Fable 5.2
API launchNew on the Anthropic API — 1,000,000-token context — $8.00 in / $40.00 out per million tokens.
-
Nov 5, 2026 · Google
Google: Gemini Omni 1.2 Flash
API launchNew on the Google API — 131,072-token context — $1.50 in / $9.00 out per million tokens.
-
Nov 4, 2026 · Google
Google: Gemini Omni 1.1 Flash
API launchNew on the Google API — 131,072-token context — $1.50 in / $9.00 out per million tokens.
-
Nov 4, 2026 · Tencent
Tencent: Hy4
API launchNew on the Tencent API — 1,048,576-token context — $0.42 in / $1.25 out per million tokens.
-
Nov 2, 2026 · Zhipu AI
Z.ai: GLM 5.3 FlashX Legal
API launchNew on the Zhipu AI API — 1,048,576-token context — $0.37 in / $1.25 out per million tokens.
Also shipped via API this month
-
Oct 26, 2026 · OpenAI
OpenAI: GPT-6.1 Terra Pro
API launchNew on the OpenAI API — 1,050,000-token context — $1.00 in / $5.00 out per million tokens.
-
Oct 26, 2026 · OpenAI
OpenAI: GPT-6.1 Terra
API launchNew on the OpenAI API — 1,050,000-token context — $1.00 in / $5.00 out per million tokens.
-
Oct 23, 2026 · Moonshot AI
MoonshotAI: Kimi K3.1 Pro
API launchNew on the Moonshot AI API — 1,048,576-token context — $0.25 in / $8.80 out per million tokens.
-
Oct 19, 2026 · MiniMax
MiniMax: MiniMax M3.1 Flash
API launchNew on the MiniMax API — 1,048,576-token context — $0.15 in / $0.60 out per million tokens.
-
Oct 8, 2026 · StepFun
StepFun: Step 5 Preview
API launchNew on the StepFun API — 1,000,000-token context — $1.00 in / $2.70 out per million tokens.
October 2026 — latest AI model releases
Newest first. Dates for API-first launches reflect first public API availability; official announcements linked where published.
-
Oct 19, 2026 · Inception
Mercury 3
Diffusion LLMWhat Inception calls the most capable diffusion-based LLM released to date, more than 40x larger than Mercury 1, with a 1M-token context window, up from 260K for Mercury 2.5. List price is $0.20 in / $0.70 out per million tokens, at 50% off ($0.10 / $0.35) during launch, on the Inception API Platform and through OpenRouter, Baseten and Vercel. API
-
Oct 17, 2026 · DeepSeek
DeepSeek-V4.1-Flash-1017
Open weightsAn update to the Sep 10 DeepSeek-V4.1-Flash that DeepSeek calls a Pareto improvement in performance and cost: stronger benchmark results, faster task completion and shorter outputs, live on its website, app and API, with new open weights. The same release opened DeepSeek's Search and Fetch APIs to everyone at $0.75 per 1,000 calls. API
-
Oct 15, 2026 · NVIDIA
Nemotron 3.5 Super VL
Open weightsAn open 120-billion-parameter model for specialized, always-on agents, trained on about 26 trillion tokens and released with open weights, datasets and training recipes on Hugging Face, OpenRouter, ModelScope and as an NVIDIA NIM. The Bolt 74B-A9B model shipped alongside it, with Bolt E4B announced for Oct 23. API
-
Oct 13, 2026 · Google DeepMind
Gemini 4 Argon
Frontier LLMGoogle's new frontier model and the first of the Gemini 4 family. It was revealed Sep 30 to cyber defenders in Google's Fairwind Program, then made globally available on Oct 13: to developers and enterprises via the API in AI Studio and the Gemini platform, to Google AI Ultra subscribers in Google Antigravity, and to Google AI Ultra and Pro subscribers in the Gemini app and AI Mode in Search, with a 1M-token unified input-output limit. Introductory pricing of $2 in / $10 out per million tokens runs until Nov 30, 2026; $4 / $20 applies after that. API
-
Oct 12, 2026 · Meta
Muse Wonder & Muse Wonder Full
Frontier LLMMeta's most capable models, built for personal and professional agentic tasks. Muse Wonder is available through the Meta Model API and Muse Code, and Muse Wonder Full is available to Muse Code subscribers.
-
Oct 7, 2026 · Anthropic
Claude Haiku 5.5
LLMThe Haiku model of the Claude 5.5 family (
claude-haiku-5-5), available on all platforms including AWS, Google Cloud and Microsoft Azure. It is priced in two tiers per million tokens: $0.10 in / $0.50 out for prompts up to 100K tokens, and $0.50 / $2.50 above that. Anthropic says it costs around 75% less to run than Haiku 4.5 on average. API -
Oct 6, 2026 · Mistral
Mistral Large 4 (public preview)
MultimodalA natively multimodal model with 1 trillion parameters, 52 billion of them active, launched as a public preview on Mistral Studio's API. API
-
Oct 6, 2026 · Google DeepMind
Nano Banana 2.1
Image genGoogle's new image model, benchmarked against Nano Banana 2 (Gemini 3.1 Flash Image) and Nano Banana Pro (Gemini 3 Pro Image), and available in Google AI Studio, the Gemini API, AI Mode in Search, Google Ads, Google Flow and Google Stitch. API
September 2026 — latest AI model releases
Newest first. Dates for API-first launches reflect first public API availability; official announcements linked where published.
-
Sep 29, 2026 · OpenAI
GPT-6.1 Sol
LLMReleased at OpenAI's DevDay for complex coding, computer use and professional work, with near-Astra performance at a lower cost than GPT-6 Astra: $2 in / $10 out per million tokens with cached input at $0.10, a 1.05M-token context window and up to 128K output tokens, text and image input, and multi-agent delegation to subagents in beta in the Responses API. API · Sol Pro
-
Sep 28, 2026 · Anthropic
Claude Sonnet 5.5
LLMThe second model in the Claude 5.5 family and a direct upgrade over Claude Sonnet 5, strongest at well-scoped everyday tasks, fixing bugs and creating documents, slides and spreadsheets; Anthropic reports 30%+ faster output and up to 30% lower cost per task than Sonnet 5 at the same $2 in / $10 out per million tokens, available on the Claude Platform, AWS, Google Cloud and Microsoft Azure. API
-
Sep 25, 2026 · Perceptron
Perceptron Mk1.5
Multimodal · EmbodiedA model built to control embodied agents: it takes text, images, video and audio, and returns text, points, boxes, polygons, clips and timestamped object tracks, with arbitrary tool calling, web search and sub-agent calls; Perceptron reports about 2–5x faster end-to-end request completion and deployments on drones, quadrupeds, smart glasses and phones, via the Perceptron Platform and SDK. API
-
Sep 22, 2026 · Anthropic
Claude Opus 5.5
Frontier LLMThe successor to Claude Opus 5 for long-running agents, coding in large codebases and professional work; Anthropic puts it at the level of Claude Fable 5.1 on most tasks while costing 40% less to run than Opus 5, at $4 in / $20 out per million tokens with a 1M-token context, in Claude apps, the Claude Platform, AWS, Google Cloud and Microsoft Foundry. API
-
Sep 22, 2026 · OpenAI
GPT-6 Sol & GPT-6 Luna
LLMThe two cheaper tiers of the GPT-6 family, below GPT-6 Astra: Sol for demanding professional and coding work at $2 in / $10 out per million tokens, Luna for high-volume, latency-sensitive tasks at $0.10 / $0.50, both with a context window of about 1M tokens in the API, Codex and ChatGPT. Sol · Sol Pro · Luna · Luna Pro
-
Sep 22, 2026 · Upstage
Solar Mini 4
Compact LLMUpstage's cost-efficient compact model for agentic use: 35B total parameters with 3B active, a 512K-token context with up to 128K output tokens, chat, reasoning, structured outputs and tool calling, and fluent Korean alongside strong English and Japanese; available on the Upstage Console and on-premises. API
-
Sep 21, 2026 · xAI
Grok 4.7
Frontier LLMxAI's most capable model for coding and knowledge work, succeeding Grok 4.6 with better self-verification on long-running software tasks and longer-context management; priced from $2 in / $6 out per million tokens, available to SuperGrok and Premium+ subscribers and through the xAI API, with invite-only red-team access for cybersecurity partners. API
-
Sep 21, 2026 · Xiaomi
MiMo-V2.6 Pro & Flash
Open weights · OmnimodalXiaomi's new flagship family: MiMo-V2.6-Pro at over 1T parameters, a Pro-UltraSpeed serving edition of the same checkpoint, and MiMo-V2.6-Flash, a 309B MoE activating 15B per token with hybrid attention — all with a 1M-token context, natively multimodal, and released under MIT with the RL checkpoints and a 9B Qwen distill on Hugging Face. Pro · UltraSpeed · Flash
-
Sep 18, 2026 · Alibaba (Qwen)
Qwen3.8-Omni-Flash
OmnimodalA natively omni-modal reasoning model that takes text, images, audio and video in one context of up to 1M tokens, and the first Qwen model built around agentic audio-video understanding and tool use; API-only at launch, with no open weights. API
-
Sep 17, 2026 · PrismML
Ternary Bonsai 2 27B
Open weights · CompressedA ternary-weight rewrite of Qwen3.8-27B that keeps 98.2% of the base model's aggregate benchmark score in a footprint about nine times smaller, so a 27B reasoning model with coding, math, tool calling and image understanding runs on one consumer GPU; Apache 2.0, 262K context. API
-
Sep 11, 2026 · Sakana AI
Fugu Max & Fugu Ultra v2
OrchestratorTwo variants of Sakana's multi-agent orchestration model: Fugu Max widens the pool of open and specialised models it can drive (including NVIDIA Nemotron) to cut token spend, while Fugu Ultra v2 targets peak quality on long multi-step coding and research tasks without relying on proprietary frontier models in its pool.
-
Sep 11, 2026 · Shanghai AI Laboratory
Atria Dawn Preview
Open weights · AgenticA 744B-parameter MoE agentic model built on the GLM-5.2 base and released under MIT with a 256K context, trained through a verifiable-experience pipeline that ties tool use to executable environments; the accompanying technical report followed on Sep 14.
-
Sep 10, 2026 · DeepSeek
DeepSeek-V4.1-Flash
Open weights · MultimodalThe first model of DeepSeek's new architecture family: a 552B-parameter MoE with a causal encoder-decoder layout that activates about 8B parameters per token on prefill and 16B on decode, native image understanding and a 1M-token context. Replaces V4 Flash and V4 Flash Vision Exp on the API as
deepseek-flash, with a price cut. API -
Sep 8, 2026 · OpenAI
GPT Image 2.5 — Flare & Sunburst
Image genTwo image models in the API and ChatGPT: Flare, the fast tier, matches or beats GPT Image 2 quality at half the latency; Sunburst, the quality tier, is built for precise edits and subject preservation in production creative work.
-
Sep 8, 2026 · Inception
Mercury 2.5
Diffusion LLMThe fastest reasoning model in production: a diffusion LLM that generates over 1,100 tokens per second, with the context window extended to 260K and tunable reasoning, native tool use and parallel tool calls added. Inception puts its intelligence level alongside cost-optimised frontier models at $0.20 in / $0.75 out per million tokens. API
-
Sep 3, 2026 · OpenAI
GPT-6 Astra
Frontier LLMOpenAI's next-generation flagship, launched in phases to a limited set of organizations first, with computer use, browsing, software engineering and cybersecurity as its headline capabilities; general API availability followed on Sep 4. API · Pro
-
Sep 2, 2026 · Google DeepMind
Gemini 3.8 Flash & 3.8 Flash Cyber
MultimodalGoogle's third Flash release in six weeks: 3.8 Flash for long-horizon software engineering and agents, plus a gated 3.8 Flash Cyber variant for security work. API
-
Sep 2, 2026 · Meta
Muse Spark 1.3
LLMAn efficiency-focused update to Muse Spark 1.2 for agentic and coding work, sustaining longer-horizon threads at the same Meta Model API price.
-
Sep 1, 2026 · Anthropic
Claude Fable 5.1 & Mythos 5.1
Frontier LLMAnthropic's coding and knowledge-work flagships; Fable 5.1 is generally available while Mythos 5.1 is the same model with relaxed safeguards, restricted to vetted cybersecurity and life-sciences organizations. API
August 2026 — latest AI model releases
Newest first. Dates for API-first launches reflect first public API availability; official announcements linked where published.
-
Aug 24, 2026 · Alibaba
Wan3.0
Video genGeneral availability on Alibaba Cloud Model Studio and Qwen Cloud after the Aug 6 public beta: clips up to 30 seconds at 1080p in a single pass (double Wan2.7's limit), with documents, spreadsheets and slide decks accepted as input alongside text, image, audio and video.
-
Aug 14, 2026 · Z.ai
GLM-5.3
LLM · Coding agentPost-training update on the GLM-5.2 base aimed at agentic coding and cyber defence; open weights followed on Hugging Face on Aug 28 after a safety review, with the smaller MIT-licensed GLM-5.3-Flash (320B-A18B) on Aug 26.
-
Aug 14, 2026 · Alibaba
Qwen3.8-27B
Open weightsDense 27B vision-language model under Apache 2.0, accepting text, images and video with a 262k native context; the small companion to Qwen3.8-Max, whose weights were published two days earlier.
-
Aug 13, 2026 · Google DeepMind
Gemini 3.7 Flash
MultimodalA coding and agent workhorse shipped three weeks after 3.6 Flash, continuing Google's fast Flash cadence.
-
Aug 12, 2026 · xAI
Grok 4.6
Frontier LLMxAI's 500k-context flagship for long-running agentic work, succeeding Grok 4.5 a month after launch.
-
Aug 12, 2026 · DeepSeek
DeepSeek-V4-Pro-0813
Open weightsV4-Pro graduates from its four-month preview: 1.57T-parameter MoE with 48B active, 1M context, 384k max output, MIT-licensed weights, with a DSpark speculative-decoding module for faster inference.
-
Aug 10, 2026 · Meta
Muse Glimmer
Open weightsMeta's return to open weights: a 30B dense multimodal model under Apache 2.0, distilled from Muse Spark and built for always-on local agents on a single consumer GPU, with a speculative-decoding drafter shipped alongside.
-
Aug 10, 2026 · OpenAI
GPT-5.6-Cyber (Daybreak Red)
Limited accessPurpose-trained cybersecurity model available only to approved Daybreak Red customers; OpenAI credits it with two previously unknown V8 vulnerabilities, patched by Google.
-
Aug 6, 2026 · OpenAI
GPT-5.6 Luna to free tier · Sol retune
Product updateLuna became the ChatGPT free-tier default, and Sol was retuned for Plus/Pro with a "thinking slider" that exposes reasoning-effort control.
-
Aug 5, 2026 · Meta
Muse Spark 1.2 & Muse Code
LLM · Coding agentSpark 1.2 brings a 1M-token context with context compaction and parallel tool calls; Muse Code, in beta, is a multi-agent coding harness with observable event logs.
-
Aug 4, 2026 · Black Forest Labs
FLUX 3 Video
Video genGeneral availability after the Jul 23 gated launch: clips up to 20 seconds with native audio and dialogue, 720p native with 1080p via upscaling.
-
Aug 4, 2026 · Liquid AI
LFM2.5-2.6B
Open weightsCompact 2.6B-parameter model aimed at on-device use.
-
Aug 3, 2026 · Alibaba
Qwen3.8-Max
Frontier LLM2.4T-parameter MoE flagship with ~95B active parameters and a 1M context window, hosted via Qwen Studio and QwenCloud at launch; open weights promised the following week.
-
Aug 3, 2026 · MiniMax
MiniMax H3
Video gen · Open weights33B omni-modal video model open-sourced under a community licence: generates video and 32 kHz stereo audio in a single pass. Note the licence excludes open-weight use in the US, EU, UK and South Korea without separate authorisation.
July 2026
Dates for API-first launches reflect first public API availability (via OpenRouter); official announcements linked where published.
-
Jul 31, 2026 · DeepSeek
DeepSeek-V4-Flash-0731
Open weightsProduction release of V4-Flash: a 284B MoE with 13B active parameters, 1.3M context, MIT licence, priced at $0.05 in / $0.16 out per million tokens.
-
Jul 24, 2026 · Anthropic
Claude Opus 5
Frontier LLMThe Claude 5 generation's heavyweight tier, succeeding Opus 4.8 at the top of the general-availability lineup.
-
Jul 21, 2026 · Google DeepMind
Gemini 3.6 Flash & 3.5 Flash Lite
MultimodalTwo fast Gemini tiers shipped the same day: 3.6 Flash for general throughput work and 3.5 Flash Lite for the cheapest per-token tier.
-
Jul 9, 2026 · OpenAI
GPT-5.6 — Sol, Luna & Terra
Frontier LLMOpenAI's three-track GPT-5.6 line (each with standard and Pro tiers) landed on the API with a 1M-token context window, after a limited preview from Jun 27.
-
Jul 8, 2026 · xAI
Grok 4.5
Frontier LLMxAI's coding-focused Grok 4 successor, shipping with a 500k-token context window.
June 2026
-
Jun 30, 2026 · Anthropic
Claude Sonnet 5
LLMThe Claude 5 generation's fast general-purpose workhorse, succeeding Sonnet 4.5 with a 1M-token context window.
-
Jun 30, 2026 · Google DeepMind
Gemini 3.1 Flash line (Nano Banana 2)
MultimodalFast Gemini 3.1 tier including the Nano Banana 2 image-generation models.
-
Jun 16, 2026 · Zhipu AI
GLM-5.2
Open weightsOpen-weights flagship that topped Hugging Face trending and runs on surprisingly modest hardware.
-
Jun 9, 2026 · Anthropic
Claude Fable 5 & Mythos 5
Frontier LLMA new Mythos-class tier above Opus. Fable 5 reached the API on Jun 9 and all Max plans on Jul 20; Mythos 5 shares the same underlying model without the added dual-use safety measures and is limited to approved organizations.
Earlier in 2026
Entries whose exact launch date we have not verified against a primary source are grouped here rather than given a date we cannot support.
-
2026 · Anthropic
Claude Opus 4.8
LLMCapstone upgrade of the Claude 4 line, adding a fast mode for quicker output at full Opus quality.
-
Apr 2026 · Alibaba
Qwen3.6
Open weightsSuccessor to Qwen3.5 (March 2026), continuing Alibaba's fast open-weights release cadence.
-
Apr 2026 · Mistral AI
Mistral Small 3.2
Open weightsCompact update to the Small line, following the Voxtral TTS release in March 2026.
-
Mar 2026 · Meta
Llama 4 Scout
Open weightsMixture-of-experts entry in the Llama 4 line.
-
Feb 2026 · Google DeepMind
Gemma 3
Open weightsOpen-weights Gemma refresh with vision capability across every size in the family.
December 2025
-
Dec 2, 2025 · Mistral AI
Mistral 3
Open weightsMistral's open-weights frontier family — Mistral Large 3 plus compact dense models — released under Apache 2.0.
November 2025
-
Nov 25, 2025 · Black Forest Labs
FLUX.2
Image genProduction-grade image generation and editing with multi-reference support and up-to-4MP output.
-
Nov 24, 2025 · Anthropic
Claude Opus 4.5
LLMReclaimed the top of the coding benchmarks at roughly a third of Opus 4.1's API price.
-
Nov 18, 2025 · Google DeepMind
Gemini 3 Pro
Frontier LLMGoogle's strongest reasoning model at launch, with Deep Think mode and generative UI in the Gemini app.
-
Nov 12, 2025 · OpenAI
GPT-5.1
LLMGPT-5 refresh with Instant and Thinking variants and adaptive reasoning depth per query.
October 2025
-
Oct 15, 2025 · Anthropic
Claude Haiku 4.5
Small LLMSmall-model tier matching Sonnet 4's coding ability at about a third of the cost and twice the speed.
-
Oct 15, 2025 · Google DeepMind
Veo 3.1
Video genVideo model update with richer audio, reference-image control, and scene extension in Flow.
September 2025
-
Sep 30, 2025 · OpenAI
Sora 2
Video genVideo generation with synchronized audio, launched alongside the invite-only Sora social app.
-
Sep 29, 2025 · Anthropic
Claude Sonnet 4.5
LLMThe mid-tier model leapfrogged Opus 4.1 on SWE-bench and sustained 30-hour autonomous agent runs.
-
Sep 29, 2025 · DeepSeek
DeepSeek-V3.2-Exp
Open weightsIntroduced DeepSeek Sparse Attention and cut API prices roughly in half.
August 2025
-
Aug 7, 2025 · OpenAI
GPT-5
Frontier LLMOpenAI's flagship: a router blending fast and reasoning models, rolled out to every ChatGPT tier including free.
-
Aug 5, 2025 · OpenAI
gpt-oss
Open weightsOpenAI's first open-weights release since GPT-2: 120B and 20B reasoning models under Apache 2.0.
-
Aug 5, 2025 · Anthropic
Claude Opus 4.1
LLMIncremental Opus upgrade focused on agentic coding, reaching 74.5% on SWE-bench Verified.
July 2025
-
Jul 9, 2025 · xAI
Grok 4
Frontier LLMFrontier model trained on the Colossus cluster, topping Humanity's Last Exam at launch.
June 2025
-
Jun 10, 2025 · Mistral AI
Magistral
ReasoningEurope's first reasoning models — open-weights Small and enterprise Medium — with transparent chain-of-thought.
May 2025
-
May 29, 2025 · Black Forest Labs
FLUX.1 Kontext
Image genIn-context image editing family: modify images with text prompts while preserving characters and style.
-
May 22, 2025 · Anthropic
Claude Opus 4 & Sonnet 4
Frontier LLMThe Claude 4 generation debut, built for long-horizon agent work and topping SWE-bench at release.
-
May 20, 2025 · Google DeepMind
Veo 3
Video genFirst mainstream video generator with native synchronized audio — dialogue, sound effects, and music.
April 2025
-
Apr 29, 2025 · Alibaba
Qwen3
Open weightsHybrid thinking/non-thinking family from 0.6B dense to 235B MoE, released under Apache 2.0.
-
Apr 16, 2025 · OpenAI
o3 & o4-mini
ReasoningReasoning models that use every ChatGPT tool — browsing, Python, vision — inside the chain of thought.
-
Apr 5, 2025 · Meta
Llama 4 Scout & Maverick
Open weightsMeta's first mixture-of-experts Llamas, natively multimodal with a 10M-token context window on Scout.
March 2025
-
Mar 25, 2025 · Google DeepMind
Gemini 2.5 Pro
Frontier LLMThinking-by-default model that took the top LMArena spot with a 1M-token context window.
February 2025
-
Feb 27, 2025 · OpenAI
GPT-4.5
LLMOpenAI's largest non-reasoning model ("Orion") — broader world knowledge and fewer hallucinations.
-
Feb 24, 2025 · Anthropic
Claude 3.7 Sonnet
LLMFirst hybrid-reasoning Claude with visible extended thinking, launched alongside Claude Code.
-
Feb 17, 2025 · xAI
Grok 3
LLMTrained on a 200k-GPU cluster, shipping Think mode and DeepSearch.
January 2025
-
Jan 20, 2025 · DeepSeek
DeepSeek-R1
Reasoning · Open weightsOpen-weights reasoning model matching OpenAI o1 at a fraction of the price — the release that triggered a $600B Nvidia sell-off.
FAQ
What is the newest AI model?
The live tracker at the top of this page is rebuilt every day from the OpenRouter models API, so it always names the newest models available to developers, with their exact first-availability dates. The month sections below carry the same releases in chronological order.
What are the latest GPT and Claude models?
OpenAI's latest flagship line is GPT-5.6 — the Sol, Luna and Terra tracks launched in July 2026, with Luna becoming the ChatGPT free-tier default in August. Anthropic's latest is the Claude 5 family — Opus 5, Fable 5, Mythos 5, and Sonnet 5 — which succeeded the Claude 4 line (Opus 4/4.1/4.5/4.8, Sonnet 4/4.5, Haiku 4.5).
How often do new AI models release?
Major labs ship a new flagship roughly every 6–12 months, with point upgrades every few weeks to months. In practice a significant model now lands almost every month: 2025 alone saw 25+ notable releases across OpenAI, Anthropic, Google, Meta, DeepSeek, Alibaba, Mistral, and xAI.
What is a frontier model?
A frontier model is one at or near the state of the art, pushing the boundary of capability in reasoning, coding, or multimodality. Current examples include Claude Fable 5, GPT-5.1, and Gemini 3 Pro. Frontier models are typically closed-weights; open-weights alternatives like DeepSeek, Qwen, and Llama usually follow within months.