Mistral 7B Instruct: benchmark results
Mistral AI's instruct fine-tune of Mistral-7B-v0.1, trained on publicly available conversation datasets and using the [INST]/[/INST] prompt format. Provider: Mistral. Released 2023-09-27. Access: Open.
Aggregate Rating 1359 ± 12, rank #1363 of 1656 rated models, from 142 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| TelcoAgent-Bench - Intent Recognition (English) | 75.5 | Intent recognition accuracy: mean cosine similarity x100 (-1 | 85.7 |
| TelcoAgent-Bench - Resolution Accuracy (English) | 87.6 | Resolution accuracy: mean cosine similarity x100 (-100 to 10 | 85.7 |
| SWE-Arena | 1002 | Elo Score | 81.9 |
| TelcoAgent-Bench - Blueprint Reliability (Arabic) | 23 | Blueprint reliability score x100 (0-100) on TelcoAgent-Bench | 71.4 |
| TelcoAgent-Bench - Blueprint Reliability (English) | 26.3 | Blueprint reliability score x100 (0-100) on TelcoAgent-Bench | 71.4 |
| TelcoAgent-Bench - Intent Recognition (Arabic) | 71.2 | Intent recognition accuracy: mean cosine similarity x100 (-1 | 71.4 |
| LogicKor - Math | 4.71 | Score (0-10) | 53.4 |
| SnakeBench | 21.4 | TrueSkill Rating | 49.4 |
| AA-Omniscience Hallucination Rate | 81.26 | Hallucination Rate (%) | 47.5 |
| LogicKor - Coding | 7 | Score (0-10) | 43.9 |
| TelcoAgent-Bench - Resolution Accuracy (Arabic) | 81.7 | Resolution accuracy: mean cosine similarity x100 (-100 to 10 | 42.9 |
| TelcoAgent-Bench - Sequence Alignment (Arabic) | 43.6 | Sequence alignment score x100 (0-100) on TelcoAgent-Bench (1 | 42.9 |
Interactive version: theaggregate.ai/model?slug=mistral-7b-instruct · How It Works · Data refreshed daily, snapshot 2026-10-20.