dolphin-2.9.1-llama-3-8B: benchmark results
Cognitive Computations' Dolphin 2.9.1, a full fine-tune of Meta's Llama-3-8B using ChatML formatting, retrained from Dolphin 2.9 with the SystemChat and Ultrachat datasets removed. Provider: Cognitive Computations. Released 2024-04-25. Access: Open.
Aggregate Rating 1341 ± 14, rank #1403 of 1656 rated models, from 44 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - GSM8K | 66.26 | Accuracy (%) (5-shot) | 81.4 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 57.24 | MC2 (%) (0-shot) | 69.6 |
| Open Korean LLM - Ko-GPQA-Diamond | 24.24 | Normalized accuracy (%) | 62.4 |
| Open LLM Leaderboard v1 - MMLU | 61.56 | Accuracy (%) (5-shot) | 52.2 |
| MT-Bench PL - Math | 4.8 | Judge Score (0-10) | 49 |
| Open Medical LLM - MedMCQA | 49.94 | Accuracy (%) | 46.9 |
| Open Medical LLM - MMLU Clinical Knowledge | 70.19 | Accuracy (%) | 45.8 |
| MT-Bench PL - Roleplay | 7.4 | Judge Score (0-10) | 44.9 |
| Open LLM Leaderboard v1 - HellaSwag | 80.55 | Normalized accuracy (%) (10-shot) | 41.3 |
| Open LLM Leaderboard v1 - ARC Challenge | 58.7 | Normalized accuracy (%) (25-shot) | 40.6 |
| MT-Bench PL - Coding | 4.6 | Judge Score (0-10) | 39.8 |
| Open Medical LLM - MMLU Medical Genetics | 71 | Accuracy (%) | 39.3 |
Interactive version: theaggregate.ai/model?slug=dolphin-2-9-1-llama-3-8b · How It Works · Data refreshed daily, snapshot 2026-10-20.