Mahou-1.2-llama3-8B: benchmark results
A DPO fine-tune released by flammen.ai of nbeerbower/llama3-KawaiiMahouSauce-8B, using flammenai/Grill-preprod-v1_chatML, flammenai/Grill-preprod-v2_chatML. The release targets casual conversation and character roleplay. Its card specifies ChatML prompting. Provider: flammen.ai. Access: Open.
Aggregate Rating 1454 ± 34, rank #3618 of 10916 rated models, from 6 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - GSM8K | 71.95 | Accuracy (%) (5-shot) | 97 |
| Open LLM Leaderboard v1 - MMLU | 68.43 | Accuracy (%) (5-shot) | 91.4 |
| Open LLM Leaderboard v1 - ARC Challenge | 69.8 | Normalized accuracy (%) (25-shot) | 84.3 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 60.5 | MC2 (%) (0-shot) | 76 |
| Open LLM Leaderboard v1 - HellaSwag | 84.65 | Normalized accuracy (%) (10-shot) | 67.6 |
| Open LLM Leaderboard v1 - WinoGrande | 77.82 | Accuracy (%) (5-shot) | 58.7 |
Interactive version: theaggregate.ai/model?slug=mahou-1-2-llama3-8b · How It Works · Data refreshed daily, snapshot 2026-11-08.