Llama-3.1-SuperNova-Lite: benchmark results
Arcee AI's Llama 3.1 8B distilled from Llama 3.1 405B Instruct logits, tuned for strong instruction following at low cost. Provider: Arcee AI. Released 2024-09-10. Access: Open.
Aggregate Rating 1516 ± 18, rank #730 of 1656 rated models, from 38 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| EVALITA - faq | 54.89 | CPS | 97.9 |
| Open CoT - LSAT Analytical Reasoning | 13.91 | CoT Gain (%) | 97.7 |
| Open LLM Leaderboard - IFEval | 80.17 | Score | 97.5 |
| Open Korean LLM Leaderboard | 49.03 | Average Score (%) | 96.4 |
| Open Ko-LLM Leaderboard | 49.03 | Average (%) | 96.3 |
| EVALITA - evalita NER | 39.49 | CPS | 95.7 |
| Open CoT - LSAT Logical Reasoning | 20.59 | CoT Gain (%) | 95 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 61.56 | Exact match, flexible extract (%) | 95 |
| Open Korean LLM - Ko-IFEval | 50.27 | Strict accuracy, prompt/instruction mean (%) | 93.7 |
| Open Korean LLM - Ko-EQ-Bench | 43.02 | EQ-Bench score | 92.7 |
| Open CoT Leaderboard | 13.97 | Average CoT Gain (%) | 91.6 |
| Open Korean LLM - KorNAT-CKA | 33.02 | Normalized accuracy (%) | 90.3 |
Interactive version: theaggregate.ai/model?slug=llama-3-1-supernova-lite · How It Works · Data refreshed daily, snapshot 2026-10-20.