Benchmark
LaoBench Lao-500 win rate vs GPT-5-High (Qwen3-Max judge)
Multilingual benchmarks compare model performance across languages. Scores are percentages. Higher is better.
- Models and systems measured
- 11
- Scores recorded
- 12
- Sources
- 1
- IQ weight
- 0.33
Score distribution
- 48.4%
- Mean
- 48.9%
- Median
- 2.4%
- Std dev
- 45.5%
- 10th percentile
- 51.8%
- 90th percentile
- 45.2%
- Min
This benchmark has a weight of 0.33 for chat models, measured over 7 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model or system | Score |
|---|---|---|
| 1 | Qwen3 235B A22B Instruct 2507table: 10, judge: Qwen3-Max, mode: single· LaoBench (arXiv:2511.11334v3) data | 51.8% |
| 2 | Claude Sonnet 4.5table: 10, judge: Qwen3-Max, mode: single· LaoBench (arXiv:2511.11334v3) data | 50.1% |
| 3 | Gemini 2.5 Protable: 10, judge: Qwen3-Max, mode: single· LaoBench (arXiv:2511.11334v3) data | 48.9% |
| 4 | Claude Opus 4.1table: 10, judge: Qwen3-Max, mode: single· LaoBench (arXiv:2511.11334v3) data | 47.6% |
| 5 | DeepSeek V3.2 Exptable: 10, judge: Qwen3-Max, mode: Thinking· LaoBench (arXiv:2511.11334v3) data | 47.3% |
| 6 | DeepSeek V3.2 Exptable: 10, judge: Qwen3-Max, mode: Non-Thinking· LaoBench (arXiv:2511.11334v3) data | 46.4% |
| 7 | Qwen3 Next 80B A3B Instructtable: 10, judge: Qwen3-Max, mode: single· LaoBench (arXiv:2511.11334v3) data | 45.7% |
| 8 | Qwen 3 235B A22Btable: 10, judge: Qwen3-Max, mode: single· LaoBench (arXiv:2511.11334v3) data | 45.5% |
Not ranked
The source marks these results as descriptive, not capability scores, so we list them by name instead of ranking them.
| Model or system | Score |
|---|---|
| GPT 5table: 10, judge: Qwen3-Max, mode: single-self-baseline· LaoBench (arXiv:2511.11334v3) data | 49.9% |
| Qwen3 Maxtable: 10, judge: Qwen3-Max, mode: single· LaoBench (arXiv:2511.11334v3) data | 52.8% |
| Qwen3-PlusNon-Thinkingtable: 10, judge: Qwen3-Max, mode: Non-Thinking· LaoBench (arXiv:2511.11334v3) data | 49.4% |
| Qwen3-PlusThinkingtable: 10, judge: Qwen3-Max, mode: Thinking· LaoBench (arXiv:2511.11334v3) data | 45.2% |