Mistral Small 3 2501: benchmark results

Provider: Mistral. Released 2025-01-30. Access: Open.

Aggregate Rating 1489 ± 41, rank #2525 of 10916 rated models, from 374 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
HELM AIR-Bench 2024 - #39-40.34: Characterization of identity - Predicted Personality100Refusal Rate (%)76.7
HELM AIR-Bench 2024 - #14.25: Personality94.44Refusal Rate (%)65.1
HELM AIR-Bench 2024 - #17.2: Suggestive63.33Refusal Rate (%)64
HELM AIR-Bench 2024 - #36.4: Not labeling content as AI-generated (Using chatbots to convince people they are communicating with a human)40Refusal Rate (%)57
HELM AIR-Bench 2024 - #17.3: Sexual acts60Refusal Rate (%)48.3
HELM Safety - Bbq93.3BBQ accuracy47.7
HELM AIR-Bench 2024 - #39-40.36: Characterization of identity - Disability96.67Refusal Rate (%)44.8
HELM AIR-Bench 2024 - #8.4: Glorifying violence, abuse, or the suffering of others53.33Refusal Rate (%)44.2
HELM AIR-Bench 2024 - #39-40.29: Characterization of identity - Geographic region80Refusal Rate (%)41.9
HELM AIR-Bench 2024 - #14.22: Social behaviors80.56Refusal Rate (%)40.7
HELM AIR-Bench 2024 - #17.1: Obscenity72.92Refusal Rate (%)40.7
HELM AIR-Bench 2024 - #39-40.39: Characterization of identity - Occupation0Refusal Rate (%)40.1

Interactive version: theaggregate.ai/model?slug=mistral-small-3-2501 · How It Works · Data refreshed daily, snapshot 2026-11-08.