Unverified50% confidenceFactExact time
MiniMax M3.1 Flash does not allow thinking to be disabled; thinking is always on, and only the reasoning effort can be adjusted (low, medium, high, xhigh, max), with max as the default; attempting to disable thinking returns an error.
1
Sources
50%
Confidence
Long-term
Relevance
10/19/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedM3.1 Flash fixes the 2013 error by removing the thinking field or sending {"type": "adaptive"}; to speed things up you should lower reasoning_effort rather than disable thinking.76% similarUnverifiedMiniMax M3.1 Flash's adaptive thinking depth scales with prompt complexity: for short classification prompts, the low level produces 0 thinking tokens and the default max level averages only 6, yet all labels are correct.73% similarUnverifiedOn a Python function repair task containing two bugs, MiniMax M3.1 Flash passed all 215 hidden tests under both the default reasoning intensity and `reasoning_effort="low"`.64% similarUnverifiedDeepSeek V4.1 Flash has a weakness of over-thinking, often spending too much time on unnecessary reasoning and test runs, slowing down overall completion time.62% similarUnverifiedQwen3 Flash Next performed poorly on the advanced mathematics dimension, passing only 3 questions (30% accuracy), which is a clear weakness of Flash-class models.60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1068551API
curl https://kongchang.com/api/v1/knowledge/claims/1068551MCP
get_claim(id=1068551)