Tess-2.0-Llama-3-8B: benchmark results
migtissera's Tess 2.0, a Llama 3 8B fine-tune on his roughly 100K-sample Tess-2.0 code and general instruction dataset, uncensored by design. Provider: migtissera. Released 2024-05-05. Access: Open.
Aggregate Rating 1411 ± 15, rank #1187 of 1656 rated models, from 24 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - MMLU | 65.97 | Accuracy (%) (5-shot) | 85.6 |
| Open Korean LLM - Ko-IFEval | 40.52 | Strict accuracy, prompt/instruction mean (%) | 83 |
| Open Korean LLM Leaderboard | 39.73 | Average Score (%) | 73.7 |
| Open Ko-LLM Leaderboard | 39.73 | Average (%) | 73.4 |
| Open Korean LLM - Ko-Winogrande | 63.22 | Accuracy (%) | 73 |
| Open Korean LLM - KorNAT-Helpful | 48.25 | Normalized accuracy (%) | 66.3 |
| Open LLM Leaderboard v1 - WinoGrande | 78.93 | Accuracy (%) (5-shot) | 66.3 |
| Open Korean LLM - Ko-EQ-Bench | 17.87 | EQ-Bench score | 64.1 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 29.49 | Exact match, flexible extract (%) | 63.9 |
| Open Korean LLM - Ko-GPQA-Diamond | 24.24 | Normalized accuracy (%) | 62.4 |
| Open LLM Leaderboard v1 - GSM8K | 47.92 | Accuracy (%) (5-shot) | 61.2 |
| Open LLM Leaderboard v1 - ARC Challenge | 61.69 | Normalized accuracy (%) (25-shot) | 53.9 |
Interactive version: theaggregate.ai/model?slug=tess-2-0-llama-3-8b · How It Works · Data refreshed daily, snapshot 2026-10-20.