Ministral-3-14B-Instruct: benchmark results
Provider: Mistral. Access: Open.
Aggregate Rating 1586 ± 25, rank #1173 of 10916 rated models, from 53 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Photonic PCFBench - Mechanism Diagnosis Field | 1 | F1 | 95 |
| Photonic PCFBench - Evidence Explanation Paired | 0.43 | BERTScore F1 | 93.3 |
| Photonic PCFBench - Geometry Description Terms | 0.61 | Term recall | 93.3 |
| Photonic PCFBench - Parameter Selection | 0.65 | Accuracy | 93.3 |
| WuYuEval - Expert Module - Elo | 1756.99 | Elo rating from pairwise judged comparisons | 81.2 |
| K12-Bench - Prerequisite Reasoning | 18.4 | Prerequisite Reasoning exact match (%, all-or-nothing over t | 66.7 |
| Photonic PCFBench - Report Completeness | 0.62 | Completeness | 66.7 |
| Photonic PCFBench - Evidence Explanation Field | 0.34 | BERTScore F1 | 65 |
| Photonic PCFBench - Pitch Prediction | 0.18 | NRMSE | 62.5 |
| WuYuEval - Expert Module - Judge Score | 0.79 | Mean anchor-calibrated LLM-as-a-judge score (0-1) | 59.4 |
| Photonic PCFBench - Parameter Modification Completeness | 1 | Completeness | 56.7 |
| Photonic PCFBench - Hole Diameter Prediction | 0.24 | NRMSE | 53.1 |
Interactive version: theaggregate.ai/model?slug=ministral-3-14b-instruct · How It Works · Data refreshed daily, snapshot 2026-11-08.