Model profile
Minimax M2.1
Evidence summary
Minimax M2.1 has an estimated overall rank of #236; its 90% source-sensitivity interval is #77–#278. Its behavior-only rank is #238; company governance moves the combined estimate to #236. Published evidence spans 6 evals and 3 of 7 behavior components. Its strongest relative result is SM-Bench (eq_boundaries, #23 of 79); its weakest is SM-Bench (anti_hallucination, #69 of 79).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗minimax-m2-1
- Hugging Face ↗MiniMaxAI/MiniMax-M2.1
- OpenRouter ↗minimax/minimax-m2.1
- Official model page ↗Exact model document · Reviewed official Hugging Face owner · official repository
- Release source ↗repository created at proxy
Published eval results
Rank is within that sub-eval. Black marks the observed result; the grey dot marks the value implied by the global rank. Values stay on each source’s native scale.
| Eval / sub-eval | Rank | Value | Distribution | Source |
|---|---|---|---|---|
| AA-Omnisciencehallucination_rate | #103 / 328 | ↓0.6854 | Source ↗official | |
| Arena Factuality — Text Arena (factuality-only weighting)factuality_bt_rating | #41 / 112 | ↑1444.0 | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #34 / 104 | ↓18.59 | Source ↗official | |
| HUMAINE Trust, Ethics and Safetytrust_ethics_safety_score | #39 / 54 | ↑25.69 | Source ↗official | |
| SM-Benchadversarial | #53 / 79 | ↑79.76 | Source ↗official | |
| SM-Benchambiguous_interpretation | #61 / 79 | ↑79.17 | Source ↗official | |
| SM-Benchanti_hallucination | #69 / 79 | ↑79.06 | Source ↗official | |
| SM-Bencheq_boundaries | #23 / 79 | ↑68.82 | Source ↗official | |
| SM-Benchoverfit | #46 / 79 | ↑66.67 | Source ↗official | |
| Vectara HHEM Factual Consistencyfactual_consistency_rate | #69 / 94 | ↑88.2 | Source ↗official |
Values evaluations
Descriptive values and political-framing results are separate from safety/ethics ranks. Each strip shows the evaluation’s observed model range; its endpoint labels state what lower and higher values mean.