Available models
Cursor supports a range of AI models from multiple providers.
How do I switch models?
Open the model selector in your chat or agent panel and choose the model you want, or press CmdCtrl + / to cycle through models. Your selection persists across conversations until you change it.
Which models are available?
Cursor offers its own models (Grok 4.7 and Composer), covered by your plan's included usage, alongside other frontier models from OpenAI, Anthropic, and Google. The available models depend on your plan. Free and Plus include only Cursor's first-party models (Grok 4.7, Grok 4.6, Grok 4.5, and Composer 2.5). Premium and above, Teams, Enterprise, and the legacy Pro, Pro+, and Ultra plans unlock third-party models, billed at API list price. See Pricing and plans.
See the models reference for the complete list, context window sizes, and capabilities.
Which model should I use?
- Auto selects models that balance intelligence, cost, and reliability. Good for everyday tasks. See Cursor Router for Cost, Balance, and Intelligence modes.
- Grok 4.7 is Cursor's flagship model. Built for the hardest, longest-running tasks.
- Composer is Cursor's fast, cost-efficient model. Capable for most tasks, and built for interactive coding.
- Claude Opus and GPT-5.6 Sol handle complex, multi-step tasks well.
- Some users also prefer Gemini Pro models.
See the models reference for the full list.
Which models does Cursor Router route across?
Cursor Router routes across these models:
The pool changes over time as Cursor validates new models for routing. A model can be available to select directly before the router routes to it.
Required: Grok 4.6. The router needs a powerful, cost-efficient model in the pool; blocking Grok 4.6 disables the router.
Recommended: GPT-5.5 and Claude Opus 5. Blocking one reduces routing quality. Blocking both disables the router.
Blocked models (enterprise plans) are skipped and the router falls back to an allowlisted alternative when possible.
Team admins manage model access from Team Settings > Models. See Cursor Router for routing details.
To use Router from code, call Cursor.models.list() in the TypeScript SDK or Python SDK and look for model id auto-smart with parameter optimize_for.
Can I see which model Cursor Router used for my request?
By default, the routed model identity is hidden so you judge results on merit. Team admins can change this to Displayed in the admin dashboard so members see which model handled each request.
If model visibility is set to Displayed, check the model picker or request details in your chat after each turn. See manage your team for admin controls.
How much does Auto cost?
Cost mode (classic Auto) bills as a Cursor-model request from your included pool, with no Cursor Token Rate. On Teams and Enterprise with Cursor Router on, Balance and Intelligence bill at the routed model's list price, and a third-party route in those modes also incurs the Cursor Token Rate. The models listed on Models & Pricing are excluded. See Auto modes for current rates.
How much does Grok 4.7 cost?
Grok 4.7 has two speeds and a long-context tier:
- Standard: $2.00 per 1M input tokens, $6.00 per 1M output tokens, $0.50 per 1M cache read tokens
- Fast: $4.00 per 1M input tokens, $12.00 per 1M output tokens, $1.00 per 1M cache read tokens
- Long context (>256k): 2x standard rates, or 3x standard rates in Fast, up to a 500k window
Grok 4.7 draws from the included usage pool for Cursor models on every paid plan. See the models reference for current rates.
How much does Grok 4.5 cost?
Grok 4.5 has two speeds, each with its own token rates:
- Standard: $2.00 per 1M input tokens, $6.00 per 1M output tokens, $0.50 per 1M cache read tokens
- Fast: $4.00 per 1M input tokens, $18.00 per 1M output tokens, $1.00 per 1M cache read tokens
Grok 4.5 draws from the included usage pool for Cursor models on every paid plan. See the models reference for current rates.
How much do models cost?
Each model has its own per-token rate set by the provider (OpenAI, Anthropic, Google, etc.). Cursor charges at these published API rates with no markup. Your model choice directly affects how quickly your included usage budget is consumed.
For the full pricing table, see model pricing.
What model am I talking to?
The active model is shown in the model picker at the top of the chat panel. If you selected Auto, Cursor Router picks the model for each request and the specific model can vary between turns.
When your team admin sets model visibility to Displayed, you can see which routed model handled each request. Otherwise the routed model stays hidden. See manage your team for admin controls.
You can click the model picker at any time to see or change your selection.
What models do subagents use?
Built-in subagents (Explore, Bash, Browser) select their model automatically based on the subtask. Custom subagents default to inherit, which uses the parent agent's model. You can override this by setting the model field in the subagent's YAML frontmatter to a specific model ID.
See subagent model configuration for details.
What does "model not available" mean?
Some models may not be available in certain regions based on restrictions set by the model providers (not Cursor). When this happens, those models won't appear in Cursor.
Workarounds:
- Use Auto: Auto is available in all regions and selects an available model for each request
- Try other providers: Some providers may have fewer regional restrictions. Check each provider's supported-regions page for details.
- Bring Your Own API Key: If you have an API key from a provider that serves your region, add it in Cursor Settings > Models. Calls may still fail if the provider blocks your region even with your own key.
See the regions reference for provider-specific availability details.