Llama3-ChatQA-1.5-8B: benchmark results
NVIDIA's Llama3-ChatQA-1.5-8B, built on Llama 3 8B with added conversational QA data for retrieval-augmented generation and tabular/arithmetic calculation, per NVIDIA's ChatQA training recipe. Provider: NVIDIA. Released 2024-04-28. Access: Open.
Aggregate Rating 1359 ± 19, rank #1364 of 1656 rated models, from 44 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open PL LLM - PSC (generative, 5-shot) | 94.44 | Binary F1 (%) | 67.9 |
| Open LLM Leaderboard v1 - MMLU | 63.76 | Accuracy (%) (5-shot) | 63.1 |
| Open LLM Leaderboard v1 - WinoGrande | 78.3 | Accuracy (%) (5-shot) | 62 |
| Open Chinese LLM - WinoGrande | 63.85 | Accuracy (%) | 61.7 |
| Open PL LLM - PolEmo2-IN (multiple choice, 5-shot) | 79.36 | Accuracy (%) | 61.3 |
| Open PL LLM - PolEmo2-IN (generative, 5-shot) | 77.56 | Accuracy (%) | 58.3 |
| Open PL LLM - Generative | 52.33 | Average Generative Score (%) | 57 |
| Open LLM Leaderboard v1 - HellaSwag | 83.17 | Normalized accuracy (%) (10-shot) | 56.5 |
| Open PL LLM - PolEmo2-OUT (generative, 5-shot) | 67.81 | Accuracy (%) | 56.5 |
| Open PL LLM - 8TAGS (multiple choice, 5-shot) | 73.49 | Accuracy (%) | 56.1 |
| Open PL LLM - PPC (generative, 5-shot) | 69.1 | Accuracy (%) | 55.8 |
| Open PL LLM - DYK (generative, 5-shot) | 52.87 | Binary F1 (%) | 54.4 |
Interactive version: theaggregate.ai/model?slug=llama3-chatqa-1-5-8b · How It Works · Data refreshed daily, snapshot 2026-10-20.