Unsupported value: 'temperature' does not support 0.7 with this model. Only the default (1) value is supported.
The message
Unsupported value: 'temperature' does not support 0.7 with this model. Only the default (1) value is supported.What it means
You set temperature on an OpenAI reasoning model. Those models only accept the default value of 1, so any other number is refused before the model runs.
What to do
Leave temperature (and top_p) out of the request for reasoning models. If an app sends it, find its temperature setting and clear it, or pick an older chat model.
The number in the middle changes with whatever you sent. This is the common one:
Unsupported value: 'temperature' does not support 0.7 with this model. Only the default (1) value is supported.Through the Python library it shows up wrapped in the SDK's own prefix, as in this agno bug report from September 2026:
Error code: 400 - {'error': {'message': "Unsupported value: 'temperature' does not support 0 with this model. Only the default (1) value is supported.What it means
Temperature used to be the first knob anyone turned: 0 for repeatable answers, around 0.7 for a bit of variety. OpenAI's reasoning models (the o-series, the GPT-5 family and GPT-6) don't take it. They accept the default of 1 and refuse anything else with this 400 before generating anything.
The same agno report shows the split cleanly. The same request with temperature=0 failed on gpt-5-mini and o4-mini, and went through on gpt-4.1-mini. That's why this usually appears the day a model name changes in a config that had carried a temperature for months.
Which settings go with it
OpenAI's migration guide for GPT-6 lists them: "When reasoning effort is not none, remove temperature, top_p, and top_logprobs. For Chat Completions, also remove logprobs." If you fix temperature and then get the same error about top_p, that's why.
The wording leaves one opening. Reasoning effort can be set to none on some models, and the guide's rule only applies when it isn't. The same page says GPT-6 Sol and Luna support none and GPT-6 Astra doesn't (Astra returns its own 400 if you try). If your app really needs a low temperature, that's the documented route on those two models; we haven't found an OpenAI page that says the same for GPT-5.
Fixing it in your code or in an app
In your own code, delete the temperature argument for these models, or only set it when the model name belongs to an older chat family. A wrapper library is often the one adding it. agno's OpenAIChat forwarded any temperature you set, and another SDK's tracker describes 26 call sites in one app failing at once after its router moved to gpt-5-mini. Their stopgap was to strip the explicit temperatures.
In a chat front end or agent tool, look for a temperature slider or a "model parameters" box and clear it (setting it to 1 also works, since 1 is the default).
If you just changed max_tokens to get here, the two errors travel together: see the max_tokens page.
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.
Unsupported value: 'temperature' does not support 0 with this model. Only the default (1) value is supported.Error code: 400 - {'error': {'message': "Unsupported value: 'temperature' does not support