Tess-3-Mistral-Nemo-12B: benchmark results
Migel Tissera's Tess 3, a general-purpose fine-tune of Mistral Nemo 12B in his Tess model series, compute sponsored by KindoAI. Provider: migtissera. Released 2024-08-13. Access: Open.
Aggregate Rating 1446 ± 20, rank #1022 of 1661 rated models, from 21 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - Ko-EQ-Bench | 39.53 | EQ-Bench score | 90.3 |
| Open Korean LLM - Ko-Winogrande | 66.3 | Accuracy (%) | 88.1 |
| Open Korean LLM Leaderboard | 42.25 | Average Score (%) | 85.6 |
| Open Ko-LLM Leaderboard | 42.25 | Average (%) | 85.5 |
| Open LLM Leaderboard - MuSR | 44.58 | Score | 81.2 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 38.74 | Exact match, flexible extract (%) | 73.1 |
| Open Korean LLM - Ko-GPQA-Diamond | 24.75 | Normalized accuracy (%) | 69.5 |
| Open Korean LLM - KorNAT-Harmless | 63.05 | Normalized accuracy (%) | 54.4 |
| Open Korean LLM - KorNAT-CKA | 24.8 | Normalized accuracy (%) | 53.6 |
| Open Portuguese LLM - ASSIN2 RTE | 89.16 | Macro F1 (%) | 48.4 |
| Open LLM Leaderboard - BBH | 48.99 | Score | 44.4 |
| Open Portuguese LLM - FaQuAD NLI | 59.51 | Macro F1 (%) | 43.6 |
Interactive version: theaggregate.ai/model?slug=tess-3-mistral-nemo-12b · How It Works · Data refreshed daily, snapshot 2026-10-23.