codex-mini-latest thinks for a long time then eventually times out on WSL

I concur, but when I use gpt4.5 or 4.0, it works fine

I get the same error on ubuntu.
● OpenAI Codex (research preview) v0.1.2505172129 │
╰──────────────────────────────────────────────────────────────╯
╭──────────────────────────────────────────────────────────────╮
│ localhost session: f533e2e8c8784ea687694c54baa34e20 │
│ ↳ workdir: ~/projects/nms │
│ ↳ model: o4-mini │
│ ↳ provider: OpenAI │
│ ↳ approval: suggest │
╰──────────────────────────────────────────────────────────────╯

system
Switched model to codex-mini-latest
user
analyze this codebase and make a document for your reference.

Interpreting user instructions

.........

Determining command usage

I’m considering using either "rg" or "cat" for reading CODEX.md. Since it's a local file, I'll run “rg CODEX.md” and remember it’s located in the root directory. The user instructed me to always use "rg" instead of "ls," but I can still explore the code tree if needed. Since they
prefer using "rg," I’ll search for the contents of CODEX.md with the command “rg -n '.*' CODEX.md.” Let’s see what I find there!

system
⚠️ Network error while contacting OpenAI. Please check your connection and try again.

system
🔗 Bug report URL: https://github.com/openai/codex/issues/new?template=2-bug-report.yml&labels=bug&version=0.1.2505172129&model=codex-mini-latest&platform=%60linux%60+%7C+%60x64%60+%7C+%606.8.0-60-generic%60&steps=-+%60%60%60%0A++analyze+this+codebase+and+make+a+document+for+your+r
eference.%0A++%60%60%60%0A++-+%601+reasoning%60+%7C+%600+tool%60

This isn't actually an issue. Call codex without first setting up your environment variable/API key. It will have you sign into your account. I think the new model is only available to Pro/Enterprise/Teams right now. After signing in the new way and granting access to the project, the new model works just fine with the CLI.

codex-mini-latest is functional in the CLI for me as well, on Ubunutu WSL2. I am a Plus sub and did the login, but that shouldn't actually matter.

According to OpenAI's announcement, the model should be available to everyone via the API. Using a Codex agent via the ChatGPT website is what's limited to higher tiers of ChatGPT subs.

For me it was thinking more than 200 sec and afte that I cancelled it.

For me it was thinking more than 200 sec and afte that I cancelled it.

I'm pretty sure this will happen if you already have an API key defined in your environment variables. You need to use --login

> For me it was thinking more than 200 sec and afte that I cancelled it. I'm pretty sure this will happen if you already have an API key defined in your environment variables. You need to use --login

I did that now and it still thinks a long time.

Im also having this problem. I am still using the API key, but I dont see how that would be the problem, since it happens every other message. I'd assume that it would never work if that was the case.
I'll switch the auth method though

I think this issue is a duplicate of https://github.com/openai/codex/issues/997. (Linking it here, so that if someone from OpenAI notices one, then they'll hopefully notice the other.)

Can people who are reporting list the exact steps to reproduce? Including what you're using on the command line? What kind of subscriber are you? I mention this, because as a pro subscriber I get a message thanking me for being one as soon as I authenticate. I'm assuming it is tied to your subscription status/API Tier somehow.

The same thing happened to me earlier, and as soon as I authenticated the appropriate way it's been working fine for me all day long.

!Image

I'm assuming it is tied to your subscription status/API Tier somehow.

As others have pointed out, sending a simple "hello" or something that does not involve tool calls will actually get a response back from codex-mini-latest. E.g., see this post. (I can confirm that this is the case.) The issue seems to be around tool calls that codex-mini tries to make.

Also, my understanding is the same as what another user posted here. In their announcement, OpenAI made the distinction between the codex-mini model available for all CLI users and the o3-derived Codex model which is available to Pro, Teams, and Enterprise users as part of their website-based coding agent.

Thank you for the feedback! Please try again on the latest version of Codex CLI. We're primarily focused on the native/Rust implementation of Codex CLI and would love to see if this issue persists there. If it does, let us know!