How to Use Grok 4.5 in Cursor ?

Grok 4.5 is built into Cursor from day one, not added as a generic external model. Cursor says the model was trained alongside the editor using trillions of tokens of developer-interaction data contributed by Cursor, which makes Cursor the best place to evaluate it on real coding workflows.

Try Apidog today

Per Cursor’s announcement, Grok 4.5 is available on all plans across desktop, web, iOS, the CLI, and the SDK. Cursor is also doubling Grok 4.5 usage for the first week after the July 8, 2026 release, so launch week is the best time to test it on real projects.

This guide shows how to enable Grok 4.5 in Cursor, how usage is counted, and when it makes sense to choose it over other models. For the model’s full spec sheet and benchmark record, see what is Grok 4.5.

What you get, by plan

Cursor includes Grok 4.5 in its first-party model pool. Individual and team plans include “significant usage” of the model, metered through the same plan usage system as Cursor’s other first-party models.

Key points:

• Launch-week usage is doubled. Through roughly July 15, 2026, Grok 4.5 usage counts at half the normal rate against your plan limits.
• Plan limits still apply. Included does not mean unlimited. Grok 4.5 still consumes your Cursor usage allowance. For more on how Cursor usage pools work, see the guides to Cursor’s rate-limit system and Ultra plan trade-offs.
• Extra usage follows API pricing passthrough. Cursor lists Grok 4.5 at $2/M input tokens and $6/M output tokens beyond plan allowances. A fast variant is listed at $4/M input and $18/M output for lower-latency serving.

How to enable Grok 4.5 in Cursor

Desktop and web

1. Update Cursor to the latest version.
2. Open the agent panel.
3. Click the model picker.
4. Select Grok 4.5.
5. If it does not appear, open Settings → Models and confirm that Grok 4.5 is enabled for your workspace.

On team plans, admins can control which models are available. If Grok 4.5 is missing from the picker, check with your workspace admin.

iOS

Use the same model picker inside the mobile app. This is useful for reviewing agent sessions or continuing a workflow away from your desk.

CLI and SDK

The Cursor CLI accepts the same first-party models as the editor. Select Grok 4.5 in the session model picker or your configuration.

If you use the Cursor SDK to script agents, select Grok 4.5 the same way you would select another first-party model. The workflow is similar to the one covered in how to use GPT-5 with Cursor CLI; only the model selection changes.

When to use Grok 4.5

Use Grok 4.5 when the task benefits from long-context, tool-heavy agent behavior.

Good fits:

• Multi-file refactors
• Long-running agent sessions
• Debugging workflows that require reading logs, tests, and source files
• Features that require planning, implementation, and iteration
• Tasks where the model needs to use Cursor tools creatively

Less ideal fits:

• Tiny edits
• Fast autocomplete-style changes
• Simple chat questions
• Low-latency tab-level tasks

A practical model-selection rule:

Use Grok 4.5 for agentic work.
Use smaller/faster models for quick edits.
Benchmark against your current default before switching permanently.

More specific trade-offs:

• Long-running agentic tasks: choose Grok 4.5. Cursor says Grok 4.5 is designed for “difficult, long-running tasks that require creatively using tools.” Cursor also reports 4.2x token efficiency against Opus 4.8, which can reduce plan usage per completed task.
• Fast tab-level edits and quick chats: choose smaller models. Grok 4.5 is fast for a flagship model, but it is not an autocomplete engine.
• Composer 2.5 still matters. Cursor’s own agent model remains in the lineup. If your workflow is tuned around it, compare both models on the same task before switching. See the Composer 2.5 breakdown for where it performs well.
• For the hardest single problems, compare frontier models. Some models in the picker still outscore Grok 4.5 on the toughest benchmarks. See Grok 4.5 vs Claude Opus 4.8 for the full comparison.

A note for EU users: Grok 4.5 is not available in the EU yet in any xAI product or the API console, with availability expected mid-July. If it does not appear in your Cursor picker, the regional rollout may be the reason.

Run a practical launch-week test

Do not evaluate Grok 4.5 based on a single prompt. Use a repeatable test while launch-week usage is doubled.

1. Pick real tasks

Choose one task from each category:

• A multi-file refactor
• A failing-test fix
• A from-scratch feature with an API surface

2. Run each task twice

For every task:

1. Start a fresh Cursor agent session.
2. Run the task with Grok 4.5.
3. Start another fresh session.
4. Run the same task with your current default model.

3. Compare the results

Track:

• How many edits you accepted
• How many edits you rejected
• Total session duration
• Number of times you had to redirect the agent
• Whether tests passed
• Whether the final implementation matched the requested behavior

A simple comparison table works well:

| Task | Model | Accepted edits | Redirects | Tests passed | Notes |
|---|---|---:|---:|---|---|
| Multi-file refactor | Grok 4.5 |  |  |  |  |
| Multi-file refactor | Current default |  |  |  |  |
| Failing-test fix | Grok 4.5 |  |  |  |  |
| Failing-test fix | Current default |  |  |  |  |
| New API feature | Grok 4.5 |  |  |  |  |
| New API feature | Current default |  |  |  |  |

If the task includes an API, verify the implementation instead of only reviewing generated code. Have the agent scaffold the endpoints, then test the contract in Apidog. Send real requests, assert response shapes, and mock unfinished endpoints so frontend work can continue in parallel.

Cursor writes the code; Apidog verifies the API behavior. The Apidog CLI runs inside Cursor, so the agent can run API test suites as part of its loop and fix failures. You can download Apidog free to wire that up.

The training-data question

Grok 4.5 learned from Cursor developer sessions, including debugging traces, multi-file diffs, and user corrections. That helps explain why it performs well in Cursor, but it is also important to understand what data was used and how privacy controls apply.

For the deeper discussion, read Grok 4.5 was trained on Cursor sessions: what that means for developers.

FAQ

Is Grok 4.5 free in Cursor?

It is included in all Cursor plans’ usage allowances, with doubled usage during launch week. xAI also says free Grok 4.5 usage is available “for a limited time” in Grok Build and Cursor. Current free paths are listed in our free-access guide.

Which Cursor plans include Grok 4.5?

All of them, per Cursor: individual and team plans across desktop, web, iOS, CLI, and SDK.

Is Grok 4.5 the default model in Cursor now?

No. It joins Cursor’s first-party model pool alongside Composer 2.5 and other models. You select it per session. It is the default in xAI’s own Grok Build.

What is the Grok 4.5 fast variant?

It is a lower-latency serving tier Cursor lists at $4/M input tokens and $18/M output tokens, compared with $2/M input and $6/M output for standard Grok 4.5. Cursor describes it as the same model served on faster infrastructure.

Does using Grok 4.5 in Cursor send my code to xAI?

Model inference runs on the provider’s infrastructure, subject to Cursor’s privacy settings and agreements. The training-data specifics are covered in our companion piece.