Models
Switch between AI models in Cursor using Chat, Composer, Tab, or Agent with different pricing tiers
With Composer, ⌘ K, and Terminal Ctrl/⌘ K, you can easily switch between different models of your choice.
Model usage
Cursor has two types of models:
Premium models
These models are usually more intelligent and count against your monthly request usage.
Free models
These models are usually faster to respond, and have unlimited usage on all our paid plans.
If you exceed your monthly premium model quota, you can enable usage based pricing to continue using these models.
For more information on model usage and quotas, see Account Usage.
Available models
Cursor has a wide range of models from a variety of providers. See the table below for a complete list of available models.
By default, the most popular models are enabled. You can add enable any of these models under Cursor Settings
> Models
.
Model | Provider | Premium | Agent | Pricing |
---|---|---|---|---|
claude-3.7-sonnet | Anthropic | ✓ | ✓ | $0.04 |
claude-3.7-sonnet MAX mode 1-4 | Anthropic | ✓ | $0.05 | |
claude-3.5-sonnet | Anthropic | ✓ | ✓ | $0.04 |
claude-3.5-haiku 5 | Anthropic | ✓ | $0.01 | |
claude-3-opus 6 | Anthropic | ✓ | $0.10 | |
cursor-small | Cursor | Free | ||
deepseek-v3 | Fireworks | Soon | Free | |
deepseek-r1 | Fireworks | ✓ | Soon | $0.04 |
gpt-4o | OpenAI | ✓ | ✓ | $0.04 |
gpt-4o-mini 7 | OpenAI | ✓ | ||
gpt-4.5-preview | OpenAI | $2.00 | ||
o1 | OpenAI | $0.40 | ||
o1-mini 6 | OpenAI | $0.10 | ||
o3-mini 5, 8 | OpenAI | ✓ | ✓ | $0.01 |
grok-2 | xAI | ✓ | $0.04 |
1 Each tool call charged like a request
2 Read file tool calls process up to 750 lines per call
3 No Agent tool call limit
4 200k max context window
5 Counts as 1/3 fast request
6 10 requests/day included on paid plan
7 Free plan gets 500 requests/day
8 Set to the high
reasoning effort
MAX mode
Models offered in MAX mode have enhanced capabilities with larger context windows and expanded reasoning.
Currently offered as an option for Claude 3.7 Sonnet, MAX mode provides a 200k token context window, unlimited Agent tool calls, and the ability to process up to 750 lines per file read operation.
When operated as an Agent, each tool call in MAX mode is charged as a separate request in addition to the initial prompt request.
Model hosting
Models are hosted on US-based infrastructure by the model’s provider, a trusted partner or Cursor.
When Privacy Mode is enabled from Settings, Cursor nor the model providers will store your data with all data deleted after each request is processed. For further details see our Privacy, Privacy Policy, and Security pages.
Model dropdown
Underneath the AI input box in both the Chat and CMD+K modes, you will see a dropdown that allows you to select the model you want to use.
For any models not visible in the dropdown, you can enable and disable them in your Cursor Settings.
Auto-select
Enabling the Auto-select switch in the model dropdown allows Cursor to proactively select the best model for your needs at that moment.
Cursor will select the premium model with the highest reliability based on current demand and best fit for the current task.
While this works best for keeping you in flow, you can also manually select a different model by disabling this option.
Thinking
Enabling the Thinking switch in the model dropdown limits the list to models that engage in more deliberate reasoning when producing responses.
Thinking models work through problems step-by-step and have deeper capacity to examine their own reasoning and correct errors.
These models often perform better on complex reasoning tasks, though they may require more time to generate their responses.
Context windows
The ‘context window’ is the amount of tokens that we provide the model with, for it to use in the conversation.
To optimize the AI performance, Cursor curates the context provided to the model to ensure the best experience.
In the Agent, the context windows is 60,000 tokens by default. For Claude 3.7 specifically, the window is up to 120,000 tokens, due to it’s better performance with longer context windows.
For Cmd-K, we limit to around 10,000 tokens to balance speed and quality.
For longer conversations, we automatically summarize the context to preserve token space. Note that these threshold are changed from time to time to optimize the experience.
Was this page helpful?