evals.report
BenchmarksLabsCompareRun guidesIn the wild
MiniMaxMiniMax M2

MiniMax M2.5

MiniMax · MiniMax M2. Released Feb 12, 2026.

MiniMax M2.5 is a model from MiniMax in the MiniMax M2 family, released Feb 12, 2026. evals.report tracks 16 reported MiniMax M2.5 benchmark scores across SWE-bench Verified, SWE-bench Pro, GPQA Diamond, Humanity's Last Exam, PostTrainBench, Epoch Capabilities Index, GDPval, SciCode, and 8 more — each shown with its benchmark, metric, source status, and date, and never combined into a single ranking.

Open16 results

Benchmark results 16

Compare this model
BenchmarkCategoryScoreMetricStatusDate
SWE-bench VerifiedCoding80.2%% resolvedVerifiedFeb 12, 2026Details
SWE-bench ProCoding55.4%% resolvedVerifiedFeb 12, 2026Details
GPQA DiamondReasoning85.2%accuracyVerifiedFeb 12, 2026Details
Humanity's Last ExamReasoning19.4%accuracyVerifiedFeb 12, 2026Details
PostTrainBenchAgents9.50%weighted average scoreOfficialFeb 12, 2026Details
Epoch Capabilities IndexReasoning147.4IndexOfficialFeb 12, 2026Details
GDPvalAgents1176EloOfficialFeb 12, 2026Details
SciCodeCoding42.6%accuracyUnverifiedFeb 12, 2026Details
Global-MMLUReasoning84.2%accuracyUnverifiedFeb 12, 2026Details
WebDev ArenaChat preference1382EloVerifiedFeb 12, 2026Details
EQ-Bench Creative Writing v3Chat preference1331EloVerifiedFeb 12, 2026Details
Design ArenaChat preference1261EloVerifiedFeb 12, 2026Details
SWE-bench MultilingualCoding68.3%% resolvedOfficialFeb 12, 2026Details
Vectara Hallucination LeaderboardOther9.1%Hallucination RateOfficialFeb 12, 2026Details
Vibe Code BenchCoding14.85%Overall accuracyVerifiedFeb 12, 2026Details
FrontierCodeCoding1.1%weighted score (Diamond)OfficialFeb 12, 2026Details