MiniMax M3.1 Flash Preview

Identity and evidence for this model. Performance depends on the endpoint, task and measurement conditions.

Updated: 2026-10-03

Availability & identity

Configured on TOKRACE; runtime availability is checked when called.

https://api.minimax.io/v1

Version and API identifier

MiniMax-M3.1-Flash-Preview

Official source · Official docs list 1M context, multimodal input and five effort levels, defaulting to max with thinking always on. Our short text test does not verify these capabilities. The exact API ID was callable on October 3, 2026, despite being absent from /models. TOKRACE trials use existing quotas and default effort. Upstream access uses subscription limits; per-token pricing is unknown and is not ranked as zero cost.

Read the first-look review and original answers

Endpoint pricing

Unknown or not billed per token. No zero-cost assumption is used.

Full price table

Standard test evidence

No published standard report yet. Task suitability has not been established.

Historical community submissions

These unverified submissions mix tasks, regions and settings; they cannot establish a fair model ranking. They are not standard-test evidence.

Endpoint	Completed / attempts	Median tok/s	Last submitted
api.minimax.io	2/4	185.4	2026-09-30

All models · Method and limitations