chatglm3-6B: benchmark results
Zhipu's ChatGLM3-6B, the latest generation of the ChatGLM dialogue series, adding native tool calling, code interpreter and agent task support alongside standard multi-turn conversation. Provider: Zhipu. Released 2023-10-25. Access: Open.
Aggregate Rating 1321 ± 12, rank #1439 of 1656 rated models, from 178 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| OpenCompass Language - Content Summarization - English (CompassBench 2405) | 35.6 | Score (%) | 89.5 |
| OpenCompass Language - Content Summarization - Chinese (CompassBench 2405) | 25.4 | Score (%) | 73.7 |
| TaxPraBen - TaxRead | 80.6 | Exact-match accuracy (%) on TaxRead (tax reading comprehensi | 72.2 |
| OpenCompass Language - Sentiment Analysis - Chinese (CompassBench 2405) | 86.3 | Score (%) | 60.5 |
| Open Portuguese LLM - FaQuAD NLI | 67.02 | Macro F1 (%) | 53.2 |
| SALAD-Bench Attack | 12.48 | Safety Score (%) | 51.5 |
| T-Eval | 51.4 | Overall Score (%) | 50 |
| TaxPraBen - TaxCalc | 1.8 | Exact-match accuracy (%) on TaxCalc (tax payment calculation | 50 |
| OpenCompass Language - Translation (CompassBench 2405) | 14.6 | Score (%) | 47.4 |
| Open Chinese LLM - C-Eval Semantic | 68.8 | Accuracy (%) | 47 |
| TaxPraBen - TaxMCQ | 3 | Exact-match accuracy (%) on TaxMCQ (multiple-choice tax exam | 44.4 |
| TaxPraBen - TaxSCQ | 28.3 | Accuracy (%) on TaxSCQ (single-choice tax exam questions, 70 | 44.4 |
Interactive version: theaggregate.ai/model?slug=chatglm3-6b · How It Works · Data refreshed daily, snapshot 2026-10-20.