New: connect Claude & other AIs to GenAIList over MCP — research the catalog and contribute to the shared knowledge base. Learn how →
AlignBench v1.1 benchmark
AI model leaderboard for the AlignBench v1.1 benchmark. Compare how large language models score on AlignBench v1.1, see the full ranking, and understand what this AI benchmark measures. Qwen3-235B-A22B currently leads with 8.94. Chinese alignment benchmark — multi-dimensional rubric judging open-ended LLM responses on Chinese tasks.
Leaderboard
| # | Model | Organization | Score | Variant | Source |
|---|---|---|---|---|---|
| #1 | Qwen3-235B-A22B | Qwen | 8.94 | thinking-v1-1 | official ↗ |
| #2 | Qwen3-235B-A22B | Qwen | 8.91 | non-thinking-v1-1 | official ↗ |
| #3 | Qwen3-32B | Qwen | 8.72 | thinking-v1-1 | official ↗ |
| #4 | Qwen3-30B-A3B | Qwen | 8.7 | thinking-v1-1 | official ↗ |
| #5 | Qwen3-32B | Qwen | 8.58 | non-thinking-v1-1 | official ↗ |
| #6 | Qwen3-14B | Qwen | 8.56 | thinking-v1-1 | official ↗ |
| #7 | Qwen3-30B-A3B | Qwen | 8.55 | non-thinking-v1-1 | official ↗ |
| #8 | Qwen3-14B | Qwen | 8.52 | non-thinking-v1-1 | official ↗ |
| #9 | Qwen3-8B | Qwen | 8.46 | thinking-v1-1 | official ↗ |
| #10 | Qwen3-8B | Qwen | 8.38 | non-thinking-v1-1 | official ↗ |
| #11 | Qwen3-4B | Qwen | 8.3 | thinking-v1-1 | official ↗ |
| #12 | Qwen2-72B | Qwen | 8.27 | instruct-cited-telechat2 | official ↗ |
| #13 | Qwen3-4B | Qwen | 8.1 | non-thinking-v1-1 | official ↗ |
| #14 | DeepSeek-V2.5 | DeepSeek | 8.04 | from-hf-readme | official ↗ |
| #15 | Telechat2-115B | China Telecom | 8.03 | from-hf-readme | official ↗ |
| #16 | DeepSeek-V2 (MoE-236B) | DeepSeek | 7.91 | cited-telechat2 | official ↗ |
| #17 | TeleChat2-35B | China Telecom | 7.88 | from-hf-readme | official ↗ |
| #18 | Qwen1.5-110B | Qwen | 7.86 | cited-telechat2 | official ↗ |
| #19 | Qwen3-1.7B | Qwen | 7.6 | thinking-v1-1 | official ↗ |
| #20 | Qwen3-1.7B | Qwen | 7.2 | non-thinking-v1-1 | official ↗ |
| #21 | TeleChat2-7B | China Telecom | 6.96 | from-hf-readme | official ↗ |
| #22 | Qwen3-0.6B | Qwen | 6.1 | thinking-v1-1 | official ↗ |
| #23 | TeleChat2-3B | China Telecom | 5.74 | from-hf-readme | official ↗ |
| #24 | Qwen3-0.6B | Qwen | 5.6 | non-thinking-v1-1 | official ↗ |
Frequently asked questions about AlignBench v1.1
What is the AlignBench v1.1 benchmark?
Chinese alignment benchmark — multi-dimensional rubric judging open-ended LLM responses on Chinese tasks.
How is the AlignBench v1.1 benchmark scored?
AlignBench v1.1 is scored using the rubric_score metric, where a higher score is better. GenAIList aggregates reported scores from model providers and papers into a single ranked leaderboard.
Which AI model scores highest on AlignBench v1.1?
As of the latest reported scores on GenAIList, Qwen3-235B-A22B achieves the highest result on AlignBench v1.1 with a score of 8.94.
Is a higher AlignBench v1.1 score better?
Yes. On AlignBench v1.1 a higher score indicates better performance, so models near the top of the leaderboard are the strongest.