Trinity-2-Codestral-22B-v0.2: benchmark results
migtissera's v0.2 fine-tune of Mistral's Codestral-22B in his Trinity series of coding-focused models. Provider: migtissera. Released 2024-08-13. Access: Open.
Aggregate Rating 1470 ± 21, rank #888 of 1654 rated models, from 20 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - Ko-IFEval | 49.22 | Strict accuracy, prompt/instruction mean (%) | 93 |
| Open Korean LLM Leaderboard | 44.76 | Average Score (%) | 90.9 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 52.46 | Exact match, flexible extract (%) | 85.6 |
| Open Korean LLM - Ko-EQ-Bench | 31.94 | EQ-Bench score | 81.2 |
| Open Portuguese LLM - FaQuAD NLI | 77.66 | Macro F1 (%) | 80.4 |
| Open LLM Leaderboard - BBH | 57.06 | Score | 78.4 |
| Open Korean LLM - KorNAT-CKA | 26.86 | Normalized accuracy (%) | 67.1 |
| Open LLM Leaderboard - GPQA | 30.79 | Score | 65.3 |
| Open Korean LLM - KorNAT-Helpful | 47.98 | Normalized accuracy (%) | 63.2 |
| Open Portuguese LLM - ASSIN2 RTE | 91.05 | Macro F1 (%) | 63.2 |
| Open Portuguese LLM - BLUEX | 55.22 | Accuracy (%) | 62.5 |
| Open Korean LLM - Ko-Winogrande | 61.17 | Accuracy (%) | 60.3 |
Interactive version: theaggregate.ai/model?slug=trinity-2-codestral-22b-v0-2 · How It Works · Data refreshed daily, snapshot 2026-10-24.