minimax-m3.1IMAGE
MiniMax M3.1 Flash API docs: reasoning_effort defaults to max and thinking can't be disabled
Before wiring it into an agent: omitting reasoning_effort defaults to max, thinking can't be turned off (disabled/none returns HTTP 400), temperature and top_p are ignored, and the benchmarks are MiniMax's own.
Lenny Enderle
@lenny_enderleArchived 9 days agoOriginally posted 2026.10.20
Lenny Enderle @lenny_enderleOriginally posted 2026.10.20
minimax m3.1 flash is all over my feed this morning. $0.15 in / $0.60 out / $0.015 cached per 1M, 1M context. before you wire it into an agent, read the api docs: - reasoning_effort defaults to max if you leave it out - thinking can't be turned off ("disabled" or "none" = http 400) - temperature / top_p are ignored so cheap only stays cheap if you set effort yourself. benchmarks are minimax's own btw. what effort level do you run flash models on?
This post quotes another post: x.com/minimax_ai/status/2112150867868541221
x.com/lenny_enderle/status/2112462534686986513Archived 2026.10.20 20:23