Mengzi3-8B-Chat: benchmark results
Langboat's Apache-2.0 8B Chinese-focused chat model (2024), built on the Llama architecture and trained on trillions of multilingual tokens. Provider: Langboat. Released 2024-09-14. Access: Open.
Aggregate Rating 1449 ± 41, rank #3829 of 10916 rated models, from 14 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Chinese LLM - GSM8K | 58.98 | Accuracy (%) | 82.2 |
| Open Chinese LLM - TruthfulQA MC | 58.96 | Accuracy (%) | 79.6 |
| Open Chinese LLM - HellaSwag | 62.6 | Accuracy (%) | 68.3 |
| Open Chinese LLM Leaderboard | 59.73 | Official Nonzero-Task Mean (%) | 66.9 |
| Open LLM Leaderboard - IFEval | 51.4 | Score | 61.3 |
| Open Chinese LLM - C-Eval Semantic | 72.72 | Accuracy (%) | 60.4 |
| Open Chinese LLM - CMMLU | 53.4 | Accuracy (%) | 50 |
| Open LLM Leaderboard - MuSR | 40.78 | Score | 48.5 |
| Open LLM Leaderboard - MATH Level 5 | 9.06 | Score | 44.4 |
| Open Chinese LLM - ARC Challenge | 50.94 | Accuracy (%) | 44.1 |
| Open LLM Leaderboard - MMLU-Pro | 31.42 | Score | 41.7 |
| Open LLM Leaderboard - BBH | 46.84 | Score | 38.6 |
Interactive version: theaggregate.ai/model?slug=mengzi3-8b-chat · How It Works · Data refreshed daily, snapshot 2026-11-08.