Ministral 3B: benchmark results
Provider: Mistral. Released 2024-10-16. Access: Open.
Unified ELO 1443 ± 1, rank #976 of 1392 rated models, from 31 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| PCB-Bench - Placement Micro QA SBERT | 58 | SBERT similarity (%) | 91.7 |
| PCB-Bench - Placement Macro QA BERTScore | 81.96 | BERTScore (%) | 75 |
| PCB-Bench - Placement Macro QA SBERT | 47.98 | SBERT similarity (%) | 75 |
| PCB-Bench - Routing Macro QA BERTScore | 82.04 | BERTScore (%) | 66.7 |
| PCB-Bench - Routing Macro QA SBERT | 49.18 | SBERT similarity (%) | 58.3 |
| PCB-Bench - Routing Micro QA SBERT | 55.52 | SBERT similarity (%) | 58.3 |
| PCB-Bench - Routing Micro QA BERTScore | 81.08 | BERTScore (%) | 50 |
| PCB-Bench - Placement Micro QA BERTScore | 80.27 | BERTScore (%) | 33.3 |
| TinyQABenchmark (core-en) | 76.9 | Exact Match | 33.3 |
| TRLawBench | 55 | Stage 1 accuracy (self-reported) | 19.4 |
| Conceptual Reasoning Index - Decision Theory (DTBench) | 19.55 | Chance-Corrected Score (0-100) | 18.8 |
| AI for Education Pedagogy - Social studies | 58.18 | Accuracy (%) | 16.6 |
Interactive version: theaggregate.ai/model?slug=ministral-3b · How It Works · Data refreshed daily, snapshot 2026-09-05.