Frontier Log

MiniMaxModel

MiniMax M2

Made by
MiniMax
Released
Announced in
MiniMax M2 & Agent: Ingenious in Simplicityminimax.io
Event
MiniMax M2
API record
/api/v1/models/minimax-m2

Benchmark results

BenchmarkScoreSource check
SWE-bench Verified69.4%
CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
CorroboratedThe same number is printed in 3 other lab documents: 1 2 3
Droid (Factory AI) harness · reported later, in MiniMax M2.1: Significantly Enhanced Multi-Language Programming, Built for Real-World Complex Tasks68.1%
CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
mini-swe-agent harness · reported later, in MiniMax M2.1: Significantly Enhanced Multi-Language Programming, Built for Real-World Complex Tasks61.0%
CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
BrowseComp44.0%
CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
CorroboratedThe same number is printed in 1 other lab document: 1
with context management · reported later, in MiniMax M2.1: Significantly Enhanced Multi-Language Programming, Built for Real-World Complex Tasks56.9%
CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
Terminal-Bench 2.0reported later, in MiniMax M2.1: Significantly Enhanced Multi-Language Programming, Built for Real-World Complex Tasks30.0%
CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
Terminal-Bench 1.046.3%
CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
Sources

Reported by MiniMax

Events

Stories

All models