Is this available via Cursor? I didn’t see it in the model list.
It is not and unlikely to be any time soon. More info here: Compatibility with DeepSeek model's design to return reasoning content after tool calls?
use deepseek v4 thinking model bug:
Request ID: ab15c5de-a582-4ca6-b28f-d5936df82be3
{“error”:“ERROR_PROVIDER_ERROR”,“details”:{“title”:“Provider Error”,“detail”:“Provider returned error: {\“error\”:{\“message\”:\“The `reasoning_content` in the thinking mode must be passed back to the API.\”,\“type\”:\“invalid_request_error\”,\“param\”:null,\“code\”:\“invalid_request_error\”}}”,“isRetryable”:false,“additionalInfo”:{},“buttons”:[],“planChoices”:[]},“isExpected”:true}
Provider Error Provider returned error: {“error”:{“message”:“The `reasoning_content` in the thinking mode must be passed back to the API.”,“type”:“invalid_request_error”,“param”:null,“code”:“invalid_request_error”}}
NLi: Provider Error Provider returned error*:* {“error”:{“message”:“The `reasoning_content` in the thinking mode must be passed back to the API.”,“type”:“invalid_request_error”,“param”:null,“code”:“invalid_request_error”}}
at Bz\_ (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:28907:24552)
at Nz\_ (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:28907:23543)
at Wz\_ (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:28908:6487)
at h6u.run (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:28908:11285)
at async vDn.runAgentLoop (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:41216:11960)
at async zkd.streamFromAgentBackend (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:41286:12151)
at async zkd.getAgentStreamResponse (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:41286:18486)
at async B3e.submitChatMaybeAbortCurrent (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:29014:16809)
at async Ma (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:40269:4230)
Created a compatibility proxy to fix this issue: https://github.com/yxlao/deepseek-cursor-proxy
The root cause is that DeepSeek API expects reasoning_content to be passed back when tool call + thinking mode is used.
It shall work for deepseek-v4-pro and deepseek-v4-flash. Let me know if it works for you!
I will try, thanks for the reference
Thank you very much, before I had a lot of error on resume, typically while waiting subagent. Or often context dropped, but now since your last version it is really fluid & usable.