neural-chat-7B-v3-3: benchmark results
Intel's neural-chat-7b-v3-3, a Mistral-7B-based chat model fine-tuned on the MetaMathQA dataset and DPO-aligned with Intel's orca_dpo_pairs, trained on Gaudi 2 hardware. Provider: Intel. Released 2023-12-09. Access: Open.
Aggregate Rating 1399 ± 24, rank #1232 of 1661 rated models, from 18 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard - MuSR | 48.6 | Score | 96.1 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 63.01 | MC2 (%) (0-shot) | 80 |
| Open LLM Leaderboard v1 - ARC Challenge | 66.89 | Normalized accuracy (%) (25-shot) | 75.1 |
| Open LLM Leaderboard v1 - GSM8K | 61.11 | Accuracy (%) (5-shot) | 73.5 |
| Open LLM Leaderboard v1 - HellaSwag | 85.26 | Normalized accuracy (%) (10-shot) | 73.1 |
| Open LLM Leaderboard v1 - WinoGrande | 79.64 | Accuracy (%) (5-shot) | 70.3 |
| Open Portuguese LLM - ASSIN2 RTE | 91.41 | Macro F1 (%) | 66.9 |
| Open Portuguese LLM - FaQuAD NLI | 71.47 | Macro F1 (%) | 63.5 |
| Open LLM Leaderboard v1 - MMLU | 63.07 | Accuracy (%) (5-shot) | 58.2 |
| Open LLM Leaderboard - IFEval | 47.63 | Score | 55.1 |
| Open Portuguese LLM - ENEM | 62.63 | Accuracy (%) | 52.5 |
| Open Portuguese LLM - BLUEX | 50.35 | Accuracy (%) | 47.2 |
Interactive version: theaggregate.ai/model?slug=neural-chat-7b-v3-3 · How It Works · Data refreshed daily, snapshot 2026-10-23.