gemini provider crashing out for most prompts

What version of Codex is running?

0.1.2504221401

Which model were you using?

gemini-2.0-flash

What platform is your computer?

linux | arm64 | 6.12.20+rpt-rpi-v8

What steps can reproduce the bug?

• ```

hello ive loaded you into an empty folder. does sound accurate?
```

• 0 reasoning | 1 tool
• ```

hello?
```

• 0 reasoning | 0 tool
• ```

yo
```

• 0 reasoning | 0 tool

What is the expected behavior?

upon prompting the model, it should have recognized the folder and responded, then offer another prompt to continue.

What do you see instead?

ai model crashed mid prompt, sending more prompts returns no response. happens pretty much every time, though sometimes it will rarely respond back a few times. specifically with gemini provider. regular openai codex with 4o-mini seems to work fine but i ran out of tokens.

Additional information

unrelated to bug report but please perhaps consider a free use open ai model such as 4.1 or 4o-mini for open ai api users, as id rather use open ai models with codex instead of gemini or other ai models, but those other ai providerss offer free tokenless models just with rate limits like gemini 2.0 flash. codex cli with gpt4.1 or something without worry of running up a massive charge on token limit as a low bank account student would be the perfect development tool imo, please consider it.