31 frontier models. All live today.

Anthropic's most capable publicly released model (Mythos-class) — next-generation intelligence for long-horizon agents, native 1M context.

The fastest Claude, near-frontier intelligence — 200k context, extended thinking.

Previous-gen flagship — 1M context, extended thinking. Consider migrating to Opus 4.7.

Anthropic's most capable general-purpose model — excels at complex reasoning and agentic coding, native 1M context.

Anthropic's newest flagship — further gains in complex reasoning and agentic coding, native 1M context.

The best balance of speed and intelligence — 1M context, extended/adaptive thinking.

The best balance of speed and intelligence — adaptive thinking on by default, native 1M context.

OpenAI's more economical coding and professional-work model — 1M context.

OpenAI's strongest mini model — good for coding, computer use, and subagents; 400k context.

OpenAI's flagship coding and professional-work model — adjustable reasoning effort, 1M context.

OpenAI's native image generation and editing (/v1/images generations + edits) — supports arbitrary resolutions.

Low-latency, cost-efficient multimodal model — good for translation, transcription, document processing, and other high-frequency lightweight tasks.

Google's most intelligent model — sustained frontier performance, excels at agentic and coding tasks.

Always points to the newest Gemini Flash generation; currently equivalent to gemini-3.5-flash.

Always points to the newest Gemini Flash-Lite generation; currently equivalent to gemini-3.1-flash-lite.

Google's flagship thinking model — excels at complex code, math, and STEM reasoning; 1M context.

Gemini's native image generation and editing (text-to-image / image+text edit) — supports 1K-4K resolutions and multiple aspect ratios.

Efficient image generation with 0.5K-4K multi-resolution output and multiple aspect ratios; integrates with search grounding.

DeepSeek's faster, more economical API — deep thinking on by default, can be disabled manually.

DeepSeek V4 Pro — significantly stronger agent capability and rich world knowledge; deep thinking on by default, can be disabled manually.

ByteDance Doubao Seed 2.0 coding model — standout frontend generation, strong multilingual support, native multimodal vision understanding.

Balances generation quality and response speed for production workloads — content creation, data analysis, and similar use cases.

Flagship all-round general model, excels at complex reasoning and long-chain task execution — multimodal understanding, long context, and tool augmentation.

Doubao coding model with precise code generation, task scheduling, and logical coordination.

Zhipu's next-gen flagship model — excels at code generation, long-horizon autonomous execution, and complex engineering optimization.

Zhipu's next-gen flagship model with native 1M-token context and adjustable thinking depth — excels at long-horizon coding and complex engineering optimization.

Moonshot Kimi K2.5 — supports text, image, and video input with thinking / non-thinking modes; strong at conversation and agent tasks.

Moonshot's strong-thinking model with multi-step tool calling and reasoning — excels at complex logical reasoning, math, and code writing.

Build complex agent harnesses — Agent Teams + Skills + Tools — for highly complex productivity tasks.

Build complex agent harnesses — Agent Teams + Skills + Tools — for highly complex productivity tasks.

MiniMax's natively multimodal, 1M-context Frontier Coding model — supports text/image/video input plus thinking.