Models
The models AutoRouter adds to Codex and Claude Code, with each one's list price and the plan it runs on.
New releases
Gemini 4 Carbon
Google's strongest model, built for its hardest problems.
Read the post· Anthropic
Claude Fable 5.2
Anthropic's frontier model for the most ambitious, open-ended technical work.
Read the post· Meta
Muse Ember
Meta's smallest model, for the hundreds of small jobs an agent runs.
Read the post· Anthropic
Claude Opus 5.5
A workhorse that's also good at visual design and animation.
Read the postAll models
| Model | Maker | Good for | Runs on | Price |
|---|---|---|---|---|
| Anthropic | Especially demanding, long-horizon coding, research and reasoning. | Claude plan | $8 / $40 | |
| OpenAI | Difficult architecture, security, science and computer-use work. | ChatGPT plan | $10 / $50 | |
| OpenAI | Default for general implementation, grounded research and analysis. | ChatGPT plan | $2 / $10 | |
| Anthropic | Broad refactoring and demanding creative or editorial work. | Claude plan | $4 / $20 | |
| Anthropic | UI implementation, writing, and a faster Claude option. | Claude plan | $2 / $10 | |
| Anthropic | Lower-priced Claude option for bounded everyday work. | Claude plan | $0.10 / $0.50 | |
| OpenAI | Lower-priced everyday coding, writing and screen or image work. | ChatGPT plan | ||
| OpenAI | A faster-turnaround alternative to GPT-6.1 Sol. | ChatGPT plan | $2 / $10 | |
| OpenAI | Routine extraction, text processing and simple tasks. | ChatGPT plan | $0.10 / $0.50 | |
| Sustained agentic engineering and complex coding. | AutoRouter Pro | |||
| Google's frontier model at a fifth of Astra's price. | AutoRouter Pro | $2 / $10 | ||
| Moonshot AI | Open-weight reasoning and coding across long supplied documents. | AutoRouter Pro | ||
| Muse Wonder | Meta | Meta's frontier model; priced above Gemini 4 Argon. | AutoRouter Pro | $4 / $20 |
| Muse Spark 1.3 | Meta | Metered choice for medium tasks when no subscription has allowance. | AutoRouter Pro | $1.25 / $4.25 |
| Muse Ember | Meta | Low-cost reasoning for agent sub-tasks and bounded coding. | AutoRouter Pro | |
| xAI | A second opinion when you ask for one. | AutoRouter Pro | ||
| Tencent Hy4 | Tencent | Long-context analysis and bounded frontend or game-development tasks. | AutoRouter Pro | |
| Qwen3.8 Max 0902 | Alibaba | Qwen alternative for bounded UI implementation and long-document code analysis. | AutoRouter Pro | |
| DeepSeek | Bounded, well-specified code changes that are easy to verify. | AutoRouter Pro | $0.30 / $1.20 | |
| Z.ai | Higher-capability open-weight coding. | AutoRouter Pro | $1.40 / $4.40 | |
| MiMo V2.6 Pro | Xiaomi | Strong open-weight generalist at a low price. | AutoRouter Pro | $0.435 / $0.87 |
| MiniMax | Low-cost open-weight option for easy tasks. | AutoRouter Pro | $0.15 / $0.60 | |
| Z.ai | Cheap structured extraction, counts and checklists. | AutoRouter Pro | $0.15 / $0.50 | |
| MiMo V2.6 Flash | Xiaomi | Cheapest open-weight option for simple tasks. | AutoRouter Pro | $0.14 / $0.28 |
| Muse Spark Open | Meta | Meta's open-weight Muse, served through Ramp. | AutoRouter Pro | $1.25 / $4.25 |
| Moonshot AI | Open-weight reasoning across long supplied documents. | AutoRouter Pro | $3 / $15 | |
| OpenAI | An older alternative kept for compatibility. | ChatGPT plan | $4 / $20 | |
| OpenAI | An older alternative kept for compatibility. | ChatGPT plan | $0.20 / $1.20 | |
| Crouton Alpha | Unnamed | Free stealth model for high-volume, repeated agent sub-tasks; its developer is unnamed. | AutoRouter Pro |
This is the set AutoRouter's team curates. You can add more models, older versions among them, in the app's Models tab. In Claude Code, Claude models run on your Claude plan. In Codex, OpenAI models run on your ChatGPT plan. Every other model comes with AutoRouter Pro and bills at the provider's price plus 10%. Prices are list prices in US dollars per million tokens, input then output, at each model's short-prompt rate: the makers' API prices for Anthropic and OpenAI, and the hosts' prices for the other models.