The model you chose is not available
The message
The model you chose is not availableWhat it means
Cursor refused to send the request to the model you picked. Recently that has mostly meant a model was broken on Cursor's side for a few hours; with your own API key it usually means the key isn't being used or can't reach that model.
What to do
Pick another model or Auto and carry on. If you use your own key, check Cursor Settings > Models: the key's toggle must be on, and the key needs access to that model.
Cursor 3.21.16 shows this in the chat when it won't run the model selected in the picker:
The model you chose is not available
Choose a different model from the picker and try again.Before 2026 the same refusal read differently, with the model's ID in the middle:
The model default does not work with your current plan or api keyThat older line was a regular on the forum. As of September 25, 2026 there are 29 topics with it in the title, from July 2025 to January 2026, and we couldn't find it in any post after February. The four topics quoting the new wording are all from September.
When a model is down on Cursor's side
Every September 2026 report we read was an outage. On September 20, people choosing Gemini 3.8 Flash High in the editor got "The model you chose is not available", even though the dropdown still listed it. A Cursor support engineer said it was fixed later that day.
On September 22 it happened to Cloud Agents. A Cursor staffer explained that "some Cloud Agents have been failing to start on Grok 4.7 with the error AI Model Not Found or The model you chose is not available", with Gemini 3.8 Flash affected too. Grok 4.7 still worked in the normal editor chat. Their workaround was Grok 4.6 or Auto, and a fix went out the same day.
So if a model you used yesterday suddenly throws this, switch to another one and check the forum's Bug Reports category or status.cursor.com. You don't need to change anything locally.
The old wording had outages too. On January 21, 2026, when people on Auto got the "model default" version, Cursor staff called it "a known issue affecting Auto mode right now" and told them to pick a specific model.
When your own API key is the reason
The "current plan or api key" half of the old message points at the other cause. Cursor lets you add your own OpenAI, Anthropic or other keys under Cursor Settings > Models, and staff have traced several of these errors to that setup:
- The key can't use that model. In December 2025 a Cursor engineer said a key "doesn't have permission to use claude-3-opus" would cause it, and the fix was a new key with access or updating the models on the existing one.
- The key isn't switched on. In August 2026 requests on Cursor 3.17.19 were rejected as
AI Model Not Foundbecause they "reached us without an attached API key". Staff said pasting the key and turning on the "Use OpenAI API key" toggle are two separate steps, and Override Base URL only applies while it's on. The user confirmed it worked after flipping the toggle and starting a new chat. - Your plan dropped to Free. Staff said in January that custom API keys only work on Pro, so if your account shows as Free after an update (staff guessed a subscription sync problem) the feature stops. Check your plan at cursor.com/settings and email hi@cursor.com if it's wrong.
One more thing we noticed: staff have said providers other than OpenAI, Anthropic and Google, connected through a custom base URL, work "as is". If you point Cursor at an unsupported provider and a model name gets rejected, that's the territory you're in.
Other lines the same feature prints
Match yours against these if the one at the top of the page is not quite it. They come from the same code and mean related things.
The model you chose is not available Choose a different model from the picker and try again.AI Model Not FoundThe model default does not work with your current plan or api keyInvalid model. The model GLM-4.7 does not work with your current plan or api keyUnknown model ID: gemini-3.8-flash