SOLAR-10.7B-Instruct-dpo: benchmark results
A fine-tune released by macadeliccc of upstage/SOLAR-10.7B-Instruct-v1.0 using Intel/orca_dpo_pairs. Its card specifies ChatML prompting. Provider: macadeliccc. Access: Open.
Aggregate Rating 1458 ± 37, rank #3430 of 10916 rated models, from 6 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 71.98 | MC2 (%) (0-shot) | 92.9 |
| Open LLM Leaderboard v1 - ARC Challenge | 71.76 | Normalized accuracy (%) (25-shot) | 90.1 |
| Open LLM Leaderboard v1 - HellaSwag | 88.08 | Normalized accuracy (%) (10-shot) | 89.4 |
| Open LLM Leaderboard v1 - MMLU | 66.06 | Accuracy (%) (5-shot) | 85.9 |
| Open LLM Leaderboard v1 - WinoGrande | 82.32 | Accuracy (%) (5-shot) | 83.5 |
| Open LLM Leaderboard v1 - GSM8K | 61.03 | Accuracy (%) (5-shot) | 73.4 |
Interactive version: theaggregate.ai/model?slug=solar-10-7b-instruct-dpo · How It Works · Data refreshed daily, snapshot 2026-11-08.