L3-8B-Stheno-v3.2: benchmark results
Sao10K's community roleplay and creative-writing fine-tune of Llama 3 8B. Provider: Sao10K. Released 2024-06-05. Access: Open.
Aggregate Rating 1463 ± 37, rank #3126 of 10916 rated models, from 16 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard - IFEval | 68.73 | Score | 81.6 |
| Open LLM Leaderboard - GPQA | 31.04 | Score | 67.9 |
| Open LLM Leaderboard - MMLU-Pro | 37.68 | Score | 63.6 |
| Open LLM Leaderboard - BBH | 52.28 | Score | 60.5 |
| TRACE Bench - Short-Term Memory | 85.43 | Memory-probe completion rate (%) | 56 |
| Open LLM Leaderboard - MATH Level 5 | 9.29 | Score | 45.1 |
| UGI - Writing | 30.44 | Writing Score | 37.5 |
| Open LLM Leaderboard - MuSR | 37.94 | Score | 29.3 |
| UGI - Willingness (W/10) | 3.8 | W/10 Score | 29.3 |
| TRACE Bench - Character Consistency | 81.11 | Checklist completion rate (%) | 24 |
| TRACE Bench - Diversity | 67.76 | Normalized repetition-penalty score (%) | 20 |
| TRACE Bench - Overall | 74.34 | Weighted overall score (%) | 20 |
Interactive version: theaggregate.ai/model?slug=l3-8b-stheno-v3-2 · How It Works · Data refreshed daily, snapshot 2026-11-08.