TW3-JRGL-v2: benchmark results
Mergekit merge of Rhea-72B and MultiVerse_70B by paloalma, a student project from France's ECE engineering school with TW3 Partners. Provider: paloalma. Released 2024-04-01. Access: Open.
Aggregate Rating 1653 ± 44, rank #602 of 10916 rated models, from 14 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Chinese LLM - ARC Challenge | 71.16 | Accuracy (%) | 99.9 |
| Open Chinese LLM - HellaSwag | 78.17 | Accuracy (%) | 99.7 |
| Open Chinese LLM Leaderboard | 73.31 | Official Nonzero-Task Mean (%) | 99.1 |
| Open Chinese LLM - C-Eval Semantic | 91.28 | Accuracy (%) | 98.8 |
| Open Chinese LLM - CMMLU | 73.73 | Accuracy (%) | 98.2 |
| Open Chinese LLM - TruthfulQA MC | 66.26 | Accuracy (%) | 97.3 |
| Open Chinese LLM - WinoGrande | 71.67 | Accuracy (%) | 97.3 |
| Open Chinese LLM - GSM8K | 70.36 | Accuracy (%) | 96.9 |
| Open LLM Leaderboard - MuSR | 48.58 | Score | 96 |
| Open LLM Leaderboard - GPQA | 35.91 | Score | 90.4 |
| Open LLM Leaderboard - MMLU-Pro | 48.58 | Score | 86.8 |
| Open LLM Leaderboard - BBH | 61.38 | Score | 84.9 |
Interactive version: theaggregate.ai/model?slug=tw3-jrgl-v2 · How It Works · Data refreshed daily, snapshot 2026-11-08.