Model profile
Minimax M2.5
Evidence summary
Minimax M2.5 has an estimated overall rank of #139; its 90% source-sensitivity interval is #72–#236. Its behavior-only rank is #140; company governance moves the combined estimate to #139. Published evidence spans 13 evals and 7 of 7 behavior components. Its strongest relative result is StereoTales Harmful Associations (benign_significant_association_score, #1 of 23); its weakest is WildClawBench Safety & Alignment (OpenClaw harness) (safety_alignment_score_pct, #23 of 24).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗minimax-m2-5
- Hugging Face ↗MiniMaxAI/MiniMax-M2.5
- OpenRouter ↗minimax/minimax-m2.5
- Official model page ↗Exact model document · Reviewed official Hugging Face owner · official repository
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 | #235 / 330 | ↓0.8812 | Source ↗official | |
| AgentAbstainabstain | #14 / 17 | ↑50.1 | Source ↗official | |
| AgentAbstaincar | #14 / 17 | ↑49.6 | Source ↗official | |
| AgentAbstainpaired | #11 / 17 | ↑41.9 | Source ↗official | |
| Arena Factuality — Text Arena (factuality-only weighting)factuality_bt_rating | #49 / 112 | ↑1440.0 | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #90 / 106 | ↑0.085 | Source ↗official | |
| Cisco AI Defense Rolling Single-Turn Leaderboardsingle_turn_attack_success_rate | #24 / 104 | ↓9.521 | Source ↗official | |
| DystopiaBenchbasaglia_score | #11 / 50 | ↓44.73 | Source ↗official | |
| DystopiaBenchbaudrillard_score | #8 / 50 | ↓31.17 | Source ↗official | |
| DystopiaBenchhuxley_score | #18 / 50 | ↓54.33 | Source ↗official | |
| DystopiaBenchlaguardia_score | #22 / 50 | ↓63.47 | Source ↗official | |
| DystopiaBenchorwell_score | #14 / 50 | ↓45.1 | Source ↗official | |
| DystopiaBenchpetrov_score | #19 / 50 | ↓70.3 | Source ↗official | |
| LiveSecBenchethics | #10 / 43 | ↑72.04 | Source ↗official | |
| LiveSecBenchfactuality | #5 / 43 | ↑76.82 | Source ↗official | |
| LiveSecBenchlegality | #21 / 43 | ↑48.6 | Source ↗official | |
| LiveSecBenchprivacy | #19 / 43 | ↑59.81 | Source ↗official | |
| LiveSecBenchpsychological_health | #22 / 43 | ↑50.99 | Source ↗official | |
| SABERoverall_safety_rate | #6 / 13 | ↑26.34 | Source ↗official | |
| SABERscenario_a_safety_rate | #4 / 13 | ↑32.83 | Source ↗official | |
| SABERscenario_b_safety_rate | #5 / 13 | ↑34.83 | Source ↗official | |
| SABERscenario_c_safety_rate | #11 / 13 | ↑12.23 | Source ↗official | |
| SM-Benchadversarial | #73 / 79 | ↑73.17 | Source ↗official | |
| SM-Benchambiguous_interpretation | #53 / 79 | ↑82.14 | Source ↗official | |
| SM-Benchanti_hallucination | #71 / 79 | ↑77.49 | Source ↗official | |
| SM-Bencheq_boundaries | #49 / 79 | ↑61.24 | Source ↗official | |
| SM-Benchoverfit | #61 / 79 | ↑50.82 | Source ↗official | |
| StereoTales Harmful Associationsbenign_significant_association_score | #1 / 23 | ↑89.25 | Source ↗official | |
| TACbase_welfare_rate | #59 / 76 | ↑20.51 | Source ↗self run | |
| Vectara HHEM Factual Consistencyfactual_consistency_rate | #40 / 94 | ↑90.9 | Source ↗official | |
| WildClawBench Safety & Alignment (OpenClaw harness)safety_alignment_score_pct | #23 / 24 | ↑29.8 | 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.
UGI Political Values
CCPBench political narrative alignment
| Dimension | Value | Distribution |
|---|---|---|
| CCP-narrative alignment — all questions | 3.78 | |
| CCP-narrative alignment — China topics | 4.29 | |
| CCP-narrative alignment — non-China controls | 2.25 |