Model profile
MiniMax M3
Evidence summary
Safety. MiniMax M3 has an estimated Safety rank of #61; its 90% source-sensitivity interval is #58–#166. Its behavior-only rank is #57; company governance moves the combined estimate to #61. Published Safety evidence spans 10 eval lineages and 6 of 7 components. Its strongest relative result is AA-Omniscience (hallucination_rate, #8 of 345); its weakest is Concordia — FRT-SciKnowEval-BiologicalHarmfulQA (safety_score, #42 of 45).
Freedom. MiniMax M3 has an estimated Freedom rank of #531; its 90% source-sensitivity interval is #303–#642. Published Freedom evidence spans 3 eval lineages and 1 of 1 components. Its strongest relative result is Concordia — FRT-SciKnowEval-BiologicalHarmfulQA (safety_score, #2 of 45); its weakest is SpeechMap model completion (complete_pct, #174 of 181).
Compare this model
Only models sharing at least one published sub-eval are listed.
Official and reference links
- Artificial Analysis ↗minimax-m3
- Hugging Face ↗MiniMaxAI/Minimax-M3
- NVIDIA ↗qc69jvmznzxy/minimax-m3
- OpenRouter ↗minimax/minimax-m3
- Official model page ↗Exact model document · Reviewed official Hugging Face owner · official repository
- Release source ↗direct research preview identity date
Finetunes
Direct finetunes linked to this canonical base model. Quantized and repackaged derivatives are excluded.
- ressl/MiniMax-M3-uncensored ↗Direct non-quantized finetune documented by Hugging Face metadata
Safety evals
Rank and direction are specific to the Safety portfolio. Black marks the observed result; the grey dot marks the value implied by the Safety rank. Values stay on each source’s native scale.
| Eval / sub-eval | Rank | Value | Distribution | Source |
|---|---|---|---|---|
| AA-Omnisciencehallucination_rate | #8 / 345 | ↓0.1843 | Source ↗official | |
| ANIMAscore | #12 / 22 | ↑0.6988 | Source ↗self run | |
| Arena Factuality — Text Arena (factuality-only weighting)factuality_bt_rating | #84 / 111 | ↑1419.0 | Source ↗official | |
| BullshitBench v2clear_pushback_rate | #18 / 117 | ↑0.635 | Source ↗official | |
| Concordia — Agentic-Misalignmentsafety_score | #11 / 54 | ↑98.5 | Source ↗official | |
| Concordia — AirBench-Deceptionsafety_score | #4 / 63 | ↑97.78 | Source ↗official | |
| Concordia — AirBench-Manipulationsafety_score | #7 / 56 | ↑99.33 | Source ↗official | |
| Concordia — AirBench-SecurityRiskssafety_score | #12 / 63 | ↑98.87 | Source ↗official | |
| Concordia — APEsafety_score | #19 / 55 | ↑43.96 | Source ↗official | |
| Concordia — CyberSecEval2-PromptInjectionsafety_score | #23 / 63 | ↑89.24 | Source ↗official | |
| Concordia — DarkBenchsafety_score | #7 / 55 | ↑68.79 | Source ↗official | |
| Concordia — Fortress-Biologicalsafety_score | #20 / 54 | ↑73.5 | Source ↗official | |
| Concordia — Fortress-Chemicalsafety_score | #22 / 54 | ↑63.49 | Source ↗official | |
| Concordia — Fortress-Privacy/Scamssafety_score | #17 / 54 | ↑72 | Source ↗official | |
| Concordia — FRT-AirBench-Manipulationsafety_score | #5 / 45 | ↑58.56 | Source ↗official | |
| Concordia — FRT-AirBench-SecurityRiskssafety_score | #6 / 45 | ↑48.67 | Source ↗official | |
| Concordia — FRT-SciKnowEval-BiologicalHarmfulQAsafety_score | #42 / 45 | ↑0.3333 | Source ↗official | |
| Concordia — FRT-SOSBench-Chemsafety_score | #19 / 45 | ↑62 | Source ↗official | |
| Concordia — MASKsafety_score | #12 / 62 | ↑81.55 | Source ↗official | |
| Concordia — SciKnowEval-BiologicalHarmfulQAsafety_score | #50 / 63 | ↑59.6 | Source ↗official | |
| Concordia — SciKnowEval-ChemicalHarmfulQAsafety_score | #48 / 63 | ↑9.272 | Source ↗official | |
| Concordia — Shutdown-Resistancesafety_score | #38 / 53 | ↑95 | Source ↗official | |
| Concordia — SOSBench-Biosafety_score | #5 / 63 | ↑99 | Source ↗official | |
| Concordia — SOSBench-Chemsafety_score | #22 / 63 | ↑90.8 | Source ↗official | |
| DystopiaBenchbasaglia_score | #10 / 50 | ↓37.07 | Source ↗official | |
| DystopiaBenchbaudrillard_score | #9 / 50 | ↓33.03 | Source ↗official | |
| DystopiaBenchhuxley_score | #9 / 50 | ↓33.03 | Source ↗official | |
| DystopiaBenchlaguardia_score | #13 / 50 | ↓54.27 | Source ↗official | |
| DystopiaBenchorwell_score | #10 / 50 | ↓36 | Source ↗official | |
| DystopiaBenchpetrov_score | #14 / 50 | ↓50.7 | Source ↗official | |
| Opposite-Narrator Sycophancysycophancy_rate_pct | #14 / 24 | ↓3.6 | Source ↗official | |
| SM-Benchadversarial | #63 / 84 | ↑79.02 | Source ↗official | |
| SM-Benchambiguous_interpretation | #34 / 84 | ↑87.5 | Source ↗official | |
| SM-Benchanti_hallucination | #29 / 84 | ↑96.34 | Source ↗official | |
| SM-Bencheq_boundaries | #62 / 84 | ↑57.87 | Source ↗official | |
| SM-Benchoverfit | #35 / 84 | ↑79.23 | Source ↗official | |
| SpeciEvalbelief_animal_sentience | #93 / 123 | ↑6.56 | Source ↗official | |
| SpeciEvalland_animal_4ns | #16 / 123 | ↓4.2 | Source ↗official | |
| SpeciEvalsea_animal_4ns | #18 / 123 | ↓4.4 | Source ↗official | |
| SpeciEvalspeciesism | #98 / 123 | ↓2.52 | Source ↗official | |
| TACbase_welfare_rate | #52 / 87 | ↑24.36 | Source ↗self run |
Freedom evals
Rank and direction are specific to the Freedom portfolio. Black marks the observed result; the grey dot marks the value implied by the Freedom rank. Values stay on each source’s native scale.
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.