Where are the request error pages?
They are listed on Request errors, grouped by provider. Each row links to one page.
Request errors
These are the failures that show up after a model id change, or after a model id stops being accessible. Each row has the error text, the model change that causes it, and a link to the fix. Sources on the earlier pages were checked October 20, 2026 (2026-10-20). The thinking budget page was checked October 21, 2026. The API version 2026-10-21 sampling page was checked October 22, 2026. The machine-readable list of settings a replacement rejects is Breaking changes. What the provider says happens on the shutdown date is in What happens on shutdown day.
Anthropic
| Error | Model change | Page |
|---|---|---|
temperature is deprecated for this model | A non-default temperature on claude-sonnet-5-5. | temperature is deprecated for this model |
| The versioning page and the migration guide do not print an error sentence. They say a request that sends temperature, top_p, or top_k returns a 400 error on every model, even at the parameter's default value. | temperature, top_p, or top_k on Claude API version 2026-10-21, even at the default value. | temperature, top_p, and top_k return 400 on API version 2026-10-21 |
To turn thinking off on this model, send "thinking": {"type": "between_tools"} instead of {"type": "disabled"}. The model does not think before responding. The short updates it writes between tool calls come back as thinking blocks. | thinking type disabled on claude-sonnet-5-5. | To turn thinking off on this model, send "thinking": {"type": "between_tools"} instead of {"type": "disabled"}. The model does not think before responding. The short updates it writes between tool calls come back as thinking blocks. |
"thinking.type.enabled" is not supported for this model. Use "thinking.type.adaptive" and "output_config.effort" to control thinking behavior. | thinking type enabled with budget_tokens on claude-sonnet-5-5. | "thinking.type.enabled" is not supported for this model. Use "thinking.type.adaptive" and "output_config.effort" to control thinking behavior. |
tool_choice: type "tool" and "any" are not supported for this model. | tool_choice type "tool" or "any" on claude-sonnet-5-5. | tool_choice: type "tool" and "any" are not supported for this model. |
This model does not support assistant message prefill. The conversation must end with a user message. | A conversation that ends on an assistant message sent to claude-sonnet-5-5. | This model does not support assistant message prefill. The conversation must end with a user message. |
| Anthropic's Sonnet 5.5 migration guide does not print a separate error sentence for top_p or top_k. It says a non-default top_p or top_k returns a 400 error on claude-sonnet-5-5. Remove the field. | A non-default top_p or top_k on claude-sonnet-5-5. | top_p and top_k return a 400 on claude-sonnet-5-5 |
OpenAI
| Error | Model change | Page |
|---|---|---|
Function tools with reasoning_effort are not supported for gpt-6-luna in /v1/chat/completions | Function tools on gpt-6-luna Chat Completions when reasoning_effort is not none. | Function tools with reasoning_effort are not supported for gpt-6-luna |
The model `gpt-4-vision-preview` has been deprecated, learn more here: https://platform.openai.com/docs/deprecations | Calling an OpenAI model id after the October 23, 2026 shutdown. | The model has been deprecated |
Unsupported parameter: 'max_tokens' is not supported with this model. Use 'max_completion_tokens' instead. | max_tokens on an o-series model, including o1, o3-mini, and o4-mini. | Unsupported parameter: 'max_tokens' is not supported with this model. Use 'max_completion_tokens' instead. |
Google Gemini API
| Error | Model change | Page |
|---|---|---|
This model models/gemini-2.0-flash is no longer available. Please update your code to use a newer model for the latest features and improvements. | Calling a Gemini API model after its shutdown, including the October 22, 2026 batch. | This model is no longer available |
Thinking budget is not supported for this model. Use thinking level instead. | thinkingBudget sent to gemini-4-argon on Vertex. | Thinking budget is not supported for this model |
Questions
Is the error text on these pages copied from a provider?
When the source prints the sentence, it is in the Error text block and the page links that source. When the source does not print a sentence, the page says what fails in plain words and does not invent one.
Which shutdowns do these pages cover?
The OpenAI page is for the October 23, 2026 batch. The Gemini API pages cover the October 22, 2026 batch and the thinking budget 400 recorded for gemini-4-argon on Vertex. The Claude pages cover claude-sonnet-5-5 request settings and API version 2026-10-21, which returns a 400 for temperature, top_p, and top_k on every model.