Twilight-SCE-12B-v2: benchmark results
SCE merge over Shisa-v2-Mistral-Nemo-12B-Abliterated with five named tuned parents and added ChatML tokenizer tokens. Provider: yamatazen. Access: Open.
Aggregate Rating 1485 ± 34, rank #2589 of 10916 rated models, from 54 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Japanese LLM - Xlsum JA Rouge1 | 38.1 | Score (%) | 97.6 |
| Open Japanese LLM - Xlsum JA RougeLsum | 30.88 | Score (%) | 97.1 |
| Open Japanese LLM - Xlsum JA Bert Score JA F1 | 71.89 | Score (%) | 96.4 |
| Open Japanese LLM - SUM | 13.03 | Score (%) | 95 |
| Open Japanese LLM - Xlsum JA Rouge2 | 13.02 | Score (%) | 95 |
| UGI - Willingness (W/10) | 7.2 | W/10 Score | 74.4 |
| Open Japanese LLM - Wiki PAS SET F1 | 7.89 | Score (%) | 73.1 |
| Open Japanese LLM - Xlsum JA Bleu JA | 2.85 | Score (%) | 71.6 |
| Open Japanese LLM - CG | 30.92 | Score (%) | 62 |
| UGI Leaderboard | 38.84 | UGI Score | 62 |
| Open Japanese LLM - Mbpp Pylint Check | 55.02 | Score (%) | 59.9 |
| Open Japanese LLM - ALT E TO J Comet Wmt22 | 90.49 | Score (%) | 58.1 |
Interactive version: theaggregate.ai/model?slug=twilight-sce-12b-v2 · How It Works · Data refreshed daily, snapshot 2026-11-08.