Mistral Vibe: Headless Token and Price Limits Now Stop Runs Before the Next Completion
Mistral VibeView original changelog
Mistral VibeHeadless --max-tokens and --max-price limits now stop the run before another root or subagent completion starts. Models priced at 0 count as free under --max-price.
Key Takeaways
--max-tokensand--max-priceare enforced before the next completion starts.- The check covers both root and subagent completions.
- Runs stop earlier instead of allowing an extra call past the limit.
- Models priced at 0 count as free under
--max-price. - It matters for scripted and CI runs with
vibe -p. - It ships in Mistral Vibe 3.1.2.
Limits That Stop Before Overspending
In Mistral Vibe 3.1.2, the headless --max-tokens and --max-price limits are checked before each new root or subagent completion begins. The run halts before another model call starts once a limit is reached, rather than allowing one more.
Free Models Under a Price Cap
Models priced at 0 now count as free under --max-price, so a price cap no longer trips on models that cost nothing.