No. 2 on the Arena text-to-image leaderboard, +79 Elo over MAI-Image-2.5 across all eight categories.

OpenAI's new flagship: far fewer made-up answers and a third of the tokens on coding-agent work, but the same overall score as GPT-5.6 Sol — at 2.5 times the price.

Compared with GPT-5.6 Sol · the model it replaces

Google's new Flash: a big jump on command-line tasks and a few points smarter, at the same price — though it burns about 30% more tokens per task, and the price doubles in 2027.

Compared with Gemini 3.7 Flash · the model it replaces

Meta's flagship edges closer to the leaders: four points up on the all-round score and clear gains on office and terminal tasks, at the same price and the same 1M-token context.

Compared with Muse Spark 1.2 · the model it replaces

Anthropic's top model, refreshed: big jumps on long coding and scientific-research tasks, and cached input now costs 75% less — same headline price as Fable 5.

Compared with Claude Fable 5 · the model it replaces

Open-weight mixture-of-experts model with 125B total parameters but only 6B active per token, a 262K native context window and a built-in vision encoder, previewing the Qwen4 architecture.

Alibaba's open 27B model, now multimodal: big gains on computer use, coding and instruction following over Qwen3.6-27B — still free to download under Apache 2.0.

Compared with Qwen3.6-27B · the model it replaces

DeepSeek's V4-Pro leaves preview with huge agentic gains — terminal, long-running coding and security tasks all jump — and stays free under the MIT license.

Compared with DeepSeek-V4-Pro (preview) · the preview it replaces

A post-training step on Grok 4.5 that adds an xhigh reasoning-effort level; cached input pricing rose from $0.30 to $0.50 per million tokens.

Open-weights release of Qwen 3.8-Max, without the cloud model's image input and non-thinking mode; a commercial licence is required above $50M in annual revenue.

Meta's first open-weight model since Llama 4: 30B parameters under Apache 2.0, able to run offline on a single 24GB consumer GPU.

xAI's new image model, built for precise editing, sharp text rendering and consistent multi-image workflows.

Introduced a contributor tier at $0.10 per million input tokens and $0.20 per million output tokens in exchange for permission to train on the user's prompts and completions.

2.4 trillion parameters with about 95 billion active per token and a 1M-token context window, served through Alibaba Cloud Model Studio.

General-availability release of the 284B-parameter V4 sibling with 13B active per token, under the MIT licence with a 1M-token context window.

Priced at $5 per million input tokens and $25 per million output tokens, unchanged from Opus 4.8, and became the default model for Claude Max subscribers.

Released alongside 3.5 Flash-Lite; Google reported up to 17% fewer tokens used than the previous Flash model.

The lowest-cost model of Google's Flash line, released the same day as 3.6 Flash.

Released without benchmarks, a model card or open weights — a break from the open Qwen-Image line.

The most capable of the three GPT-5.6 models (Sol, Terra and Luna), priced at $5 per million input tokens and $30 per million output tokens.

The mid tier of the GPT-5.6 trio (Sol / Terra / Luna), positioned by OpenAI as the balanced everyday model: $2 per million input tokens and $12 per million output tokens after the 30 July price cut.

The lightest tier of the GPT-5.6 trio, positioned by OpenAI as the fast, affordable model: $0.20 per million input tokens and $1.20 per million output tokens after the 30 July price cut.

Opened the Meta Model API to outside developers in public preview, priced at $1.25 per million input tokens and $4.25 per million output tokens.

Added configurable reasoning effort (low, medium, high) with a 500K-token context window, priced at $2 per million input tokens and $6 per million output tokens.

Meta's first in-house image model, free in the Meta AI app, Instagram Stories and WhatsApp.

Apache 2.0 model for Lean 4 proof engineering with 119B parameters and 6B active, solving 587 of 672 PutnamBench problems.

Replaced Sonnet 4.6 as the default model on Anthropic's Free and Pro plans, priced at $2 per million input tokens and $10 per million output tokens.

Google's first open-weight text diffusion model: 26B parameters with 3.8B active, Apache 2.0, generating over 1,000 tokens per second on a single H100.

Anthropic's new top-end model: a big jump on hard coding tasks over Opus 4.8 — at twice the price.

Compared with Claude Opus 4.8 · the model it replaces

Microsoft's first in-house coding model — small, cheap, and ahead of Claude Haiku 4.5 on the coding tests Microsoft published.

Compared with Claude Haiku 4.5 · a competitor — comparison chosen by Microsoft

Launched at No. 2 for image editing on Arena, unveiled at Build 2026.

Alibaba's proprietary multimodal agent model with a 1M-token context, generally available API-only on Alibaba Cloud Model Studio at $0.40/$1.60 per 1M tokens.

Anthropic's updated top-end model: steady coding gains over Opus 4.7 at the same price, plus a new Fast mode that runs the model at a third of the cost.

Compared with Claude Opus 4.7 · the model it replaces

Google's lightweight Flash model: it edges out the bigger Gemini 3.1 Pro on a coding test, while running far more token-efficient than the Flash it replaces — at the same price.

Compared with Gemini 3 Flash · the model it replaces

Google's new any-to-video model: text, images, audio or video in, physics-aware video out, editable in plain English — 10-second clips, free in YouTube Shorts and the Gemini app.

Alibaba's new top Qwen model: clear gains on terminal coding work and hard reasoning tests over 3.6-Plus, plus a much larger memory — at the same price.

Compared with Qwen 3.6-Plus · the model it replaces

OpenAI's new lightweight model: far fewer made-up answers and much stronger math than GPT-5.3 Instant Mini — and it's now the default free model in ChatGPT.

Compared with GPT-5.3 Instant Mini · the model it replaces

xAI's update to Grok 4.20: it can now hold a whole book of text at once, watch video, and costs 40% less for the text you send in — with a few points higher on a broad skills test.

Compared with Grok 4.20 · the model it replaces

DeepSeek's new open flagship: about 8 times more text in memory than V3.2, while using a fraction of the computing power — and it solves about 8 in 10 real-world coding tasks.

Compared with DeepSeek-V3.2 · the version it replaces

OpenAI's GPT-5.5 roughly doubles its reasoning over very long documents and writes shorter answers than GPT-5.4 — but it costs twice as much to run.

Compared with GPT-5.4 · the model it replaces

The first open-weight variant of the Qwen3.6 family — a coding-focused 27B dense model under Apache 2.0 with a 262K native context, extensible to 1M.

Reasons before it draws: up to 2K output, up to eight coherent images per prompt, and reliable text rendering including non-Latin scripts.

Anthropic's updated flagship: a clear jump on real-world coding and much sharper at reading charts than 4.6 — at the same price.

Compared with Claude Opus 4.6 · the model it replaces

Unified architecture for image generation and editing; the video models followed the week after, with select models back under Apache 2.0.

An open-weight (Apache 2.0) code agent for the Lean 4 proof assistant that generates code and formally proves implementations against specifications.

Google's updated flagship: its score on brand-new reasoning puzzles more than doubled over Gemini 3 Pro — at the same price.

Compared with Gemini 3 Pro · the model it replaces

xAI's new flagship Grok: the first version that retrains itself every week from real-world feedback, and it makes up fewer answers than any model tested so far.

Compared with Grok 4.1 · the version it replaces

Unifies generation and editing in a single 7B model — nearly 3× smaller than Qwen-Image, yet ahead on every major benchmark reported.

Anthropic's Opus 4.6: a big jump on brand-new reasoning puzzles and the first Opus that can read 1M tokens of text at once — but everyday coding is unchanged.

Compared with Claude Opus 4.5 · the model it replaces

An open-weight (Apache 2.0) coding-agent model with 80B total and 3B active parameters and a 256K context, built for coding agents and local development.

xAI's image-and-video generator: 10-second 720p clips with native audio, including lip-synced dialogue.

Image inputs and outputs 20% cheaper in the API than GPT Image 1.

Shipped as a closed API only — no weights, a break from the open 2.1 and 2.2.

Built on Gemini 3 Pro: far better text rendering inside images, consistent identity across up to five subjects, and 2K/4K output.

Adds 4K and native portrait output on top of Veo 3's synchronized audio.

Microsoft's first in-house image model, debuting in the top 10 on LMArena and inside Bing Image Creator.

A smaller GPT Image that costs 80% less in the API.

Adds synchronized sound and 'cameos' (insert yourself into clips), launched as a social app. OpenAI shut the Sora app down on April 26, 2026 and ends the API on September 24, 2026.

Google's image generation moves into Gemini itself: consistent characters across edits and multi-image blending, at Flash speed.

Alibaba's open image model (Apache 2.0), strong at rendering text inside images.

Open weights again under Apache 2.0 — the last fully open Wan video release.

Google's video model gains synchronized audio — dialogue, ambient sound and effects generated with the footage.

Alibaba's open-weight video model, released under Apache 2.0.