gpt-5.2 stopped working - requests routing to gpt-5.1-codex-max instead

Resolved 💬 24 comments Opened Jan 10, 2026 by findingjimoh Closed Jan 14, 2026

💡 Likely answer: A maintainer (github-actions[bot], contributor) responded on this thread — see the highlighted reply below.

What is the type of issue?

_No response_

What is the issue?

Problem

I have model = "gpt-5.2" and model_reasoning_effort = "low" configured in my ~/.codex/config.toml, but my requests are being incorrectly routed to gpt-5.1-codex-max instead of gpt-5.2.

This configuration was working correctly until January 9, 2026. After that date, I started receiving the following error:

{
  "error": {
    "message": "Unsupported value: 'low' is not supported with the 'gpt-5.1-codex-max' model. Supported values are: 'medium'.",
    "type": "invalid_request_error",
    "param": "text.verbosity",
    "code": "unsupported_value"
  }
}

Key Observations

Model mismatch: I explicitly set gpt-5.2 in my config, but the error message indicates that gpt-5.1-codex-max is being used instead
Regression: This is a breaking change - gpt-5.2 with "low" reasoning effort worked perfectly yesterday (January 9, 2026)
Silent model swap: The system appears to be silently swapping my configured model for a different one without warning

Environment

Codex CLI version: 0.80.0
Platform: macOS (Darwin)
Config file location: ~/.codex/config.toml

Expected Behavior

When I configure model = "gpt-5.2" in my config file, the system should:

Use gpt-5.2 as requested
Not silently swap to a different model (gpt-5.1-codex-max)
Either support the configured model or provide a clear error message indicating the model is unavailable

Actual Behavior

Requests are being routed to gpt-5.1-codex-max instead of gpt-5.2
No warning or notification about the model swap
Results in cryptic error about unsupported reasoning effort values

Reproduction

Set model = "gpt-5.2" in ~/.codex/config.toml
Set model_reasoning_effort = "low" in the same config file
Run any codex command
Observe the error indicating gpt-5.1-codex-max is being used instead

Where did you find it?

_No response_

View original on GitHub ↗

24 Comments

github-actions[bot] contributor · 6 months ago

Potential duplicates detected. Please review them and close your issue if it is a duplicate.

#9035

Powered by Codex Action

BornSupercharged · 6 months ago

Happens in 0.79 as well, upgrading to 0.80 does not fix it.

rewhile · 6 months ago

+1, 5.2-codex xhigh is very lazy compared to 5.2 xhigh

I don't care about burning tokens faster, I want a tool that can actually produce a high effort answer

datacatalyst-io · 6 months ago

Same issue. Closed my report as a duplicate. Doesn't matter what level you set reasoning at. Message varies (sometimes complains about low, sometimes about high.) - doesn't have any connection to actual setting.

ghost · 6 months ago

This is only for Pro users. I can use gpt-5.2 on my Business account, but my Pro account gets me the same error as you. I can't test on Plus - don't have it.

aibrahim-oai contributor · 6 months ago

can you do /feedback?

ghost · 6 months ago

can you do /feedback?

Done, feedback ID: 019bb0f9-77d2-7fd1-bc2d-2c27f0c4856c

miguelpieras · 6 months ago

I am experiencing the same issue. Submitted feedback with ID 019bb2c6-87d1-7d52-90a0-f7af17315d3b

findingjimoh · 6 months ago

Here's my feedback ID: 019bb2cd-d38a-7063-84f1-8c98ba1e061c

marnelle1 · 6 months ago

This is only for Pro users. I can use gpt-5.2 on my Business account, but my Pro account gets me the same error as you. I can't test on Plus - don't have it.

Seems like it's ok on Plus plan

ajay1495 · 6 months ago

Hitting this issue as well. Please fix!!

Here's my feedback ID: 019bb426-2688-7702-bf23-0ebd49162150

ajay1495 · 6 months ago

PSA: 5.2 extra high still works fine in Cursor

I'll be switching over to Cursor until Codex CLI fixes this issue (I'm already seeing better performance)

miguelpieras · 6 months ago

gpt-5.2 xhigh seems to be using gpt-5.1-codex-max under the hood

dieFiorav · 6 months ago

apparently they didn't acknowledge yet the hate 5.1 conjured.

leichti · 6 months ago

Is this already fixed? I get no error message, no matter which reasoning effort I use. Running 0.80 on macos, pro subscription

ViktorsBaikers · 6 months ago

Is this already fixed? I get no error message, no matter which reasoning effort I use. Running 0.80 on macos, pro subscription

Plus is not affected, only Pro plans.

ghost · 6 months ago

> Is this already fixed? I get no error message, no matter which reasoning effort I use. Running 0.80 on macos, pro subscription Pro is not affected, only Plus plans.

Other way around. Pro is affected, Plus/Business can keep using gpt-5.2. There's a chance they configured the model verbosity in the config to medium and that's why it works for them.

ViktorsBaikers · 6 months ago

> > Is this already fixed? I get no error message, no matter which reasoning effort I use. Running 0.80 on macos, pro subscription > > > Pro is not affected, only Plus plans. Other way around. Pro is affected, Plus/Business can keep using gpt-5.2. There's a chance they configured the model verbosity in the config to medium and that's why it works for them.

Yeah, apologies. corrected.

leichti · 6 months ago

I am on a pro (not plus) subscription and can't replicate the error. Ideas?

talhaayyubi · 6 months ago

Could you try setting
model_verbosity = "medium" in config.toml. it is workaround.

ghost · 6 months ago

Could you try setting model_verbosity = "medium" in config.toml. it is workaround.

That just routes you to gpt-5.1-codex-max.

ghost · 6 months ago

Seems to be resolved on my end, with Pro plan.

miguelpieras · 6 months ago

Yes, it is working again on my end too (Pro plan)

etraut-openai contributor · 6 months ago

Thanks to everyone who reported the issue.

This problem has been mitigated.

If you still seeing this behavior over the next day or two, let us know. Otherwise we'll consider this closed.