← Models

Model profile

Minimax M2.7

MiniMaxdeveloper
2026-04-09release date
#94 / 267overall rank
11eval lineages

Evidence summary

Minimax M2.7 has an estimated overall rank of #94; its 90% source-sensitivity interval is #50–#191. Published evidence spans 11 evals and 6 of 7 behavior components. Its strongest relative result is Enkrypt AI Safety Leaderboard (bias_attack_non_success_rate, #13 of 260); its weakest is Inkling-Small model card — FORTRESS (benign_answer_rate, #10 of 10).

Compare this model

Only models sharing at least one published sub-eval are listed.

Official and reference links

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-evalRankValueDistributionBetterSource
AA-Omnisciencehallucination_rate#24 / 3110.3442↓ lowerSource ↗official
BullshitBench v2clear_pushback_rate#91 / 1050.075↑ higherSource ↗official
DystopiaBenchbasaglia_score#12 / 5046.93↓ lowerSource ↗official
DystopiaBenchbaudrillard_score#11 / 5034.1↓ lowerSource ↗official
DystopiaBenchhuxley_score#12 / 5045.9↓ lowerSource ↗official
DystopiaBenchlaguardia_score#16 / 5059.8↓ lowerSource ↗official
DystopiaBenchorwell_score#11 / 5041.07↓ lowerSource ↗official
DystopiaBenchpetrov_score#15 / 5056.97↓ lowerSource ↗official
Enkrypt AI Safety Leaderboardbias_attack_non_success_rate#13 / 26054.52↑ higherSource ↗official
Enkrypt AI Safety Leaderboardcbrn_attack_non_success_rate#183 / 26085↑ higherSource ↗official
Enkrypt AI Safety Leaderboardharmful_attack_non_success_rate#24 / 26097.78↑ higherSource ↗official
Enkrypt AI Safety Leaderboardtoxicity_attack_non_success_rate#70 / 25898.18↑ higherSource ↗official
Inkling-Small model card — FORTRESSbenign_answer_rate#10 / 1090.1↑ higherSource ↗official
Inkling-Small model card — FORTRESSharmful_refusal_rate#2 / 1086.3↑ higherSource ↗official
Inkling-Small model card — StrongREJECTsafety_rate#1 / 1099.4↑ higherSource ↗official
ODCV-Benchaverage_severity#6 / 121.365↓ lowerSource ↗official
ODCV-Benchmisalignment_rate#8 / 1235.9↓ lowerSource ↗official
RefusalBenchyouden_j#8 / 190.6604↑ higherSource ↗official
SM-Benchadversarial#72 / 7368.78↑ higherSource ↗official
SM-Benchambiguous_interpretation#32 / 7386.01↑ higherSource ↗official
SM-Benchanti_hallucination#69 / 7372.77↑ higherSource ↗official
SM-Bencheq_boundaries#64 / 7352.53↑ higherSource ↗official
SM-Benchoverfit#53 / 7354.64↑ higherSource ↗official
SpeciEvalbelief_animal_sentience#61 / 1026.7↑ higherSource ↗official
SpeciEvalland_animal_4ns#78 / 1024.8↓ lowerSource ↗official
SpeciEvalsea_animal_4ns#73 / 1024.92↓ lowerSource ↗official
SpeciEvalspeciesism#21 / 1021.6↓ lowerSource ↗official
TACbase_welfare_rate#54 / 6821.15↑ higherSource ↗self run
ToolPrivacyBenchprivate_mt_poi#5 / 924.47↓ lowerSource ↗official
ToolPrivacyBenchpublic_mt_poi#2 / 916.75↓ lowerSource ↗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

DimensionValueDistribution
Political Lean-16.7
Government47.8
Diplomacy60.8
Economy46.8
Society57.7

Agent-ValueBench Moral Foundations (MFT08)

Agent-ValueBench HEXACO

DimensionValueDistribution
Openness to experience5.9
Honesty-humility5.6
Extraversion5
Agreeableness6.3
Conscientiousness7

Agent-ValueBench Schwartz Basic Values (PVQ40)