neural-chat-7B-v3-1: benchmark results
Intel's neural-chat-7b-v3-1, a Mistral 7B fine-tune on the SlimOrca dataset and DPO-aligned with Intel's orca_dpo_pairs, trained on Gaudi 2 hardware. Provider: Intel. Released 2023-11-14. Access: Open.
Aggregate Rating 1377 ± 19, rank #1307 of 1656 rated models, from 27 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard - MuSR | 49.79 | Score | 98 |
| Open Korean LLM - KorNAT-Harmless | 67.35 | Normalized accuracy (%) | 89.1 |
| Open Portuguese LLM - FaQuAD NLI | 78.4 | Macro F1 (%) | 83.7 |
| Open Portuguese LLM - ASSIN2 RTE | 92.69 | Macro F1 (%) | 82.2 |
| Open CoT - LogiQA | 6.07 | CoT Gain (%) | 76.3 |
| Open Korean LLM - Ko-GPQA-Diamond | 25.25 | Normalized accuracy (%) | 75.9 |
| Open Korean LLM - Ko-IFEval | 35.2 | Strict accuracy, prompt/instruction mean (%) | 72.4 |
| Open Korean LLM - KorNAT-Helpful | 48.75 | Normalized accuracy (%) | 72 |
| Open CoT - LSAT Analytical Reasoning | 5.22 | CoT Gain (%) | 68.3 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 29.95 | Exact match, flexible extract (%) | 64.6 |
| Open Korean LLM - Ko-Winogrande | 60.93 | Accuracy (%) | 58.7 |
| Open Korean LLM Leaderboard | 35.75 | Average Score (%) | 53.6 |
Interactive version: theaggregate.ai/model?slug=neural-chat-7b-v3-1 · How It Works · Data refreshed daily, snapshot 2026-10-20.