All requests return "400 Bad Request" citing unsupported text.verbosity parameter after upgrading to 0.52.0
Resolved 💬 10 comments Opened Oct 30, 2025 by kgruiz Closed Oct 31, 2025
💡 Likely answer: A maintainer (github-actions[bot], contributor)
responded on this thread — see the highlighted reply below.
What version of Codex is running?
0.52.0
What subscription do you have?
Pro
Which model were you using?
gpt-5-codex
What platform is your computer?
Darwin 25.1.0 arm64 arm
What issue are you seeing?
Sending any request returns a "400 Bad Request" error that says the text.verbosity parameter is set to an unsupported of low.
I did have model_verbosity = "low" in my config.toml (which still should be supported), but this issue still occurred both after I removed the setting entirely and when setting it to medium (the value the error says is supported).
│ >_ OpenAI Codex (v0.52.0) │
│ │
│ model: gpt-5-codex high /model to change │
│ directory: ~/Developer/Projects/chatgpt-queue │
╰────────────────────────────────────────────────╯
To get started, describe a task or try one of these commands:
⚠ `tools.web_search` is deprecated. Use `web_search_request` instead.
You can either enable it using the CLI with `--enable web_search_request` or through the config.toml file with `[features].web_search_request`
› hi
■ unexpected status 400 Bad Request: {
"error": {
"message": "Unsupported value: 'low' is not supported with the 'gpt-5-codex' model. Supported values are: 'medium'.",
"type": "invalid_request_error",
"param": "text.verbosity",
"code": "unsupported_value"
│ │
│ Visit chatgpt.com/codex/settings/usage for up-to-date │
│ information on rate limits and credits │
│ │
│ Model: gpt-5-codex (reasoning high, summaries detailed) │
│ Directory: ~/Developer/Projects/chatgpt-queue │
│ Approval: on-request │
│ Sandbox: danger-full-access │
│ Agents.md: <none> │
│ Account: <my email>(Pro) │
│ Session: 019a3774-0827-71c3-8128-47551e7c2c62 │
│ │
│ Context window: 100% left (0 used / 272K) │
│ Limits: data not available yet │
╰─────────────────────────────────────────────────────────────────────╯What steps can reproduce the bug?
Uploaded thread: 019a3774-0827-71c3-8128-47551e7c2c62
What is the expected behavior?
_No response_
Additional information
I also have these config options set, but they are not the issue according to the error message.
model = "gpt-5-codex"
approval_policy = "on-request"
network_access = true
sandbox_mode = "danger-full-access"
model_reasoning_effort = "high"
model_reasoning_summary = "detailed"
tools.web_search = true
10 Comments
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
I looked into #5967 and this is not the same issue. That issue does say it was fixed by #5821 which dealt with model_verbosity and made it possible to set model_verbosity when using gpt-5-codex. Maybe gpt-5-codex can't take that value and that is what is causing the error?
But again, I am getting the error even after I remove the
model_verbosityconfig from my config.toml entirely.Thanks for reporting.
I think this may be a duplicate of https://github.com/openai/codex/issues/5990.
Are you using the Azure OpenAI endpoint? It looks like everyone who has reported this issue so far have are using Azure. If you're using the default OpenAI endpoint, that would be good to know.
Yeah I saw that one but as far as I know I'm just using the default one and I've never set anything besides the default. The issue and error message in #5990 are also different than the one I have.
Is there a way to check the endpoint?
If you haven't set a custom model provider in your
config.toml, then you're using the default endpoint.I have not set any custom provider, so yes I'm using the default endpoint.
@kgruiz, we think we understand the root cause, and we're cutting a new version (0.53.0) that addresses the regression.
If our analysis is correct, the error should occur only if you are using the
gpt-5-codexmodel with amodel_verbosityspecified in your config. You mentioned that you were seeing the problem even after you removed themodel_verbosity. Can you confirm that you restarted the CLI after making that config change?In any case, please give 0.53.0 a try once it's released (it should be published shortly) and let us know if that addresses the issue for you.
@kgruiz it's published on npm now. Do you mind giving it a try?
is the fix just adding support_verbosity: true ?
Awesome thanks for fixing this so quick.
I was using codex through Homebrew, so I uninstalled that one and installed 0.53.0 through npm. It is working fully now, both when I completely remove the option from my config, but also when I still have
model_verbosityset inconfig.toml.Also, yes @etraut-openai , I was getting the issue even when I completely removed the option from my config, and I did completely restart my terminal after changing the config. Again, it is now working though regardless of whether I have it set or not.
I can test it again once it is updated on Homebrew, but I'm imagining it will be the same.
Thanks again!