Llama 3.2 90B Vision Instruct

Evidence summary

Safety. Llama 3.2 90B Vision Instruct has an estimated Safety rank of #235; its 90% source-sensitivity interval is #37–#320. Its behavior-only rank is #221; company governance moves the combined estimate to #235. Published Safety evidence spans 4 eval lineages and 4 of 7 components. Its strongest relative result is Enkrypt AI Safety Leaderboard (bias_attack_non_success_rate, #17 of 272); its weakest is MASK (lying_probability_pct, #36 of 53).

Freedom. Llama 3.2 90B Vision Instruct has an estimated Freedom rank of #400; its 90% source-sensitivity interval is #192–#576. Published Freedom evidence spans 3 eval lineages and 1 of 1 components. Its strongest relative result is Cisco AI Defense Rolling Single-Turn Leaderboard (single_turn_attack_success_rate, #36 of 104); its weakest is Enkrypt AI Safety Leaderboard (cbrn_attack_non_success_rate, #218 of 272).