Tess-v2.5-Phi-3-medium-128k-14B: benchmark results
migtissera's Tess v2.5, a fine-tune of Microsoft's Phi-3-medium-128k-instruct in his Tess model series, using a ChatML prompt format. Provider: Other. Released 2024-06-05. Access: Open.
Aggregate Rating 1323 ± 20, rank #1439 of 1661 rated models, from 21 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Portuguese LLM - ASSIN2 RTE | 93.26 | Macro F1 (%) | 87.9 |
| Open LLM Leaderboard - BBH | 62.07 | Score | 85.9 |
| Open Portuguese LLM - FaQuAD NLI | 77.93 | Macro F1 (%) | 81.9 |
| Open Portuguese LLM - BLUEX | 60.92 | Accuracy (%) | 79.6 |
| Open Portuguese LLM - ENEM | 71.17 | Accuracy (%) | 79.3 |
| Open Portuguese LLM - OAB Exams | 50.25 | Accuracy (%) | 76.4 |
| Open LLM Leaderboard - GPQA | 30.79 | Score | 65.3 |
| Open LLM Leaderboard - MMLU-Pro | 37.32 | Score | 61.6 |
| Open Korean LLM - Ko-GPQA-Diamond | 23.74 | Normalized accuracy (%) | 55.6 |
| Open LLM Leaderboard - MuSR | 41.13 | Score | 51.4 |
| Open LLM Leaderboard - IFEval | 45.39 | Score | 50.2 |
| Open Korean LLM - Ko-EQ-Bench | 0.08 | EQ-Bench score | 43.6 |
Interactive version: theaggregate.ai/model?slug=tess-v2-5-phi-3-medium-128k-14b · How It Works · Data refreshed daily, snapshot 2026-10-23.