Gemini 3.0 Pro - Out Now!

After a long-awaited return to the list of frontier models, Gemini is back with Gemini 3.0 Pro!

Gemini 3.0 Pro is high on reasoning and intelligence, with strong performance across agentic coding and visual / frontend design tasks!

As with the previous Gemini models, it remains a 1M token limit model, as is the most expensive Gemini model yet, priced slightly above the previous 2.5 Pro pricing.

For the exact pricing and model capabilities, see our documentation here:

Let us know how you get on with this new model :tada:

Has anyone tried planning yet with this model? I did ask quite big change involving lots of new things in UI, but I got very small plan with very generic highlights only.

Screenshot 2025-11-18 at 11.43.38 AM can’t use it

Switching to Gemini in an existing agent discussion did not work. Try creating a new agent discussion, it seems to work fine.


not sure if this is on cursor’s end but the last dot in gemini 3 response is on a new line, happens sometimes

Any thoughts on this model compared to Claude 4.5 Sonnet?

It is going to kill Sonnet 4.5 in next 2 weeks. Could happen faster, but it takes some time before people realize how good it is compared

Not working, 400 error code :
Request ID: 24f00ae3-0263-4afc-828c-c8f4b550d700
{“error”:“ERROR_OPENAI”,“details”:{“title”:“Unable to reach the model provider”,“detail”:“We encountered an issue when using your API key: Provider was unable to process your request\n\nAPI Error:\n\n\nRequest failed with status code 400: [{\n \"error\": {\n \"code\": 400,\n \"message\": \"Invalid role: model\",\n \"status\": \"INVALID_ARGUMENT\"\n }\n}\n]\n”,“additionalInfo”:{},“buttons”:,“planChoices”:},“isExpected”:true}
ConnectError: [invalid_argument] Error
at QJc.$endAiConnectTransportReportError (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:4989:399336)
at Ywo._doInvokeHandler (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:492:36007)
at Ywo._invokeHandler (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:492:35749)
at Ywo._receiveRequest (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:492:34514)
at Ywo._receiveOneMessage (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:492:33336)
at GDt.value (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:492:31429)
at _e._deliver (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:49:2962)
at _e.fire (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:49:3283)
at ggt.fire (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:4974:12156)
at MessagePort. (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:7498:18439)

Constant crashes, the API is over the limit, the world is totally at it. Let’s give it a few days to calm down a bit, meanwhile, Cursor team - great work!

Hey Cursor team, whenever a new model like this is released, do you guys need a few days to fine tune the system prompt and tool harness? This would be good to know if its the case, as then we won’t write off new models right away if we experience poor performance with them on launch day.

Yes, we’ve tuned it beforehand! However, the model going out on launch day usually varies from the early tests (this is standard for all model providers)

I’ve been using Cursor for a year, and in terms of cost and output quality, I think I’m currently getting the best results with Gemini 3

Getting the same error

The API key gives an error: invalid role model. The API key is very important for me and the cursor users.

Request ID: 09055a4e-a855-4522-a088-1a1550cc0aaa
{“error”:“ERROR_OPENAI”,“details”:{“title”:“Unable to reach the model provider”,“detail”:“We encountered an issue when using your API key: Provider was unable to process your request\n\nAPI Error:\n\n\nRequest failed with status code 400: [{\n \"error\": {\n \"code\": 400,\n \"message\": \"Invalid role: model\",\n \"status\": \"INVALID_ARGUMENT\"\n }\n}\n]\n”,“additionalInfo”:{},“buttons”:,“planChoices”:},“isExpected”:true}
ConnectError: [invalid_argument] Error
at QJc.$endAiConnectTransportReportError (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:4989:399336)
at Ywo._doInvokeHandler (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:492:36007)
at Ywo._invokeHandler (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:492:35749)
at Ywo._receiveRequest (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:492:34514)
at Ywo._receiveOneMessage (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:492:33336)
at GDt.value (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:492:31429)
at _e._deliver (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:49:2962)
at _e.fire (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:49:3283)
at ggt.fire (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:4974:12156)
at MessagePort. (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:7498:18439)

Issue should be fixed now! Let us know if it work as intended

Any news on when Gemini 3.0 Pro High will be available?

I don’t see it in my Cursor IDE. Is it available just for the API at this moment?

I am getting this error as well, but only through the cursor-agent CLI. Requests through the IDE appear to be going through without issue for me.

I am still getting the “Error reaching model provider” error through cursor-agent (but not the IDE).