Blossom-V6-14B: benchmark results
A conversational fine-tune released by Azure99 based on Qwen/Qwen2.5-14B, using Azure99/blossom-v6-sft-stage1, Azure99/blossom-v6-sft-stage2. The card specifies one epoch for stage 1 and three epochs for stage 2. Provider: Azure99. Access: Open.
Aggregate Rating 1542 ± 32, rank #1807 of 10916 rated models, from 56 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard - MATH Level 5 | 52.57 | Score | 97.9 |
| Open Japanese LLM - Xlsum JA Bert Score JA F1 | 70.56 | Score (%) | 87.7 |
| Open Japanese LLM - SUM | 11.34 | Score (%) | 86.4 |
| Open Japanese LLM - Xlsum JA Rouge2 | 11.34 | Score (%) | 86.4 |
| Open LLM Leaderboard - MMLU-Pro | 45.44 | Score | 84.4 |
| Open Japanese LLM - Xlsum JA RougeLsum | 25.6 | Score (%) | 83.4 |
| Open Japanese LLM - Wiki Dependency SET F1 | 38.52 | Score (%) | 80.5 |
| Open Japanese LLM - Jcommonsenseqa Exact Match | 95.62 | Score (%) | 79.1 |
| Open Japanese LLM - Xlsum JA Rouge1 | 29.66 | Score (%) | 78.8 |
| Open Japanese LLM - FA | 27.73 | Score (%) | 76.6 |
| Open LLM Leaderboard - IFEval | 63.95 | Score | 76.4 |
| Open Japanese LLM - Wiki NER SET F1 | 9.73 | Score (%) | 76.1 |
Interactive version: theaggregate.ai/model?slug=blossom-v6-14b · How It Works · Data refreshed daily, snapshot 2026-11-08.