Rankings / MiniMax
Minimax M2 1
released 2025-12-23
BenchAtlas Index
as of 2026-07-14
base configuration
rank #184
9 families · 5 categories · high
base configuration
rank #211
9 families · 4 categories · medium
Benchmark evidence
20 results
agentic coding
| τ²-Bench | 85.4 % | independent | Artificial Analysis |
coding
| LiveCodeBench v6 implementation=artificial-analysis | 81.0 % | independent | Artificial Analysis | |
| SciCode | 40.7 % | independent | Artificial Analysis |
external indices
| Intelligence Index v4.1 | 31.4 points | independent | Artificial Analysis | |
| AA Math Index | 82.7 points | independent | Artificial Analysis |
knowledge science
| GPQA Diamond implementation=artificial-analysis | 83.0 % | independent | Artificial Analysis | |
| GPQA Diamond implementation=vals-ai | 78.5 % | independent | Vals AI | |
| Humanity's Last Exam implementation=artificial-analysis | 22.2 % | independent | Artificial Analysis | |
| MMLU-Pro implementation=artificial-analysis | 87.5 % | independent | Artificial Analysis | |
| MMLU-Pro implementation=vals-ai | 87.0 % | independent | Vals AI |
long context instruction
| AA-LCR | 59.0 % | independent | Artificial Analysis | |
| IFBench | 69.9 % | independent | Artificial Analysis |
professional
| CaseLaw | 55.8 % | independent | Vals AI | |
| CorpFin | 58.9 % | independent | Vals AI | |
| LegalBench | 75.4 % | independent | Vals AI | |
| MedQA | 91.2 % | independent | Vals AI | |
| TaxEval | 66.4 % | independent | Vals AI |
reasoning math
| AIME year=2025 · implementation=artificial-analysis | 82.7 % | independent | Artificial Analysis | |
| AIME implementation=vals-ai | 77.9 % | independent | Vals AI | |
| MATH-500 implementation=vals-ai | 89.0 % | independent | Vals AI |
Agent + model results
systems, not bare-model scores
| agent + model Crux + Minimax M2 1 | Terminal-Bench 2.0 | 36.6 % | community | Terminal-Bench Leaderboard |
| agent + model terminus-2 + Minimax M2 1 | Terminal-Bench 2.0 | 29.2 % | community | Terminal-Bench Leaderboard |
| agent + model Artificial Analysis harness + Minimax M2 1 | Terminal-Bench Hard | 28.8 % | independent | Artificial Analysis |
These scores measure the whole agent system (scaffold, tools, budgets) — they are never merged into the bare model’s numbers.