Benchmark
VITA-Bench
Vision benchmarks measure how well models understand images. Scores are percentages. Higher is better.
- Models measured
- 29
- Scores recorded
- 42
- Sources
- 3
- IQ weight
- 1.05
Score distribution
- 27.1%
- Mean
- 26.3%
- Median
- 13.2%
- Std dev
- 12.3%
- 10th percentile
- 45.6%
- 90th percentile
- 0.8%
- Min
This benchmark has a weight of 1.05 for chat models, measured over 29 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model | Score |
|---|---|---|
| 1 | Qwen3.5 397B A17B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 49.7% |
| 2 | Qwen3.5 397B A17BQwen3.5 release/model card · published setting· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 49.7% |
| 3 | Qwen3.7 Max· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 47.9% |
| 4 | Qwen3.7 MaxQwen3.7-Max release · subdomain average · Claude 4.5 Sonnet judge· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 47.9% |
| 5 | Qwen3.7 Plus· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 45.6% |
| 6 | Qwen3.7 PlusQwen3.7-Plus release · subdomain average · Claude 4.5 Sonnet judge· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 45.6% |
| 7 | Qwen3.6 Plus· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 44.3% |
| 8 | Qwen3.6 PlusQwen official release table · thinking mode unless benchmark requires otherwise · average subdomain score · Claude Sonnet 4 judge· Reported by Alibaba Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 44.3% |
| 9 | Qwen3.5 27B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 41.9% |
| 10 | Qwen3 Max· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 40.9% |
| 11 | Qwen3.6 35B A3B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 35.6% |
| 12 | Qwen3.6 FlashQwen3.6-35B-A3B official model card · hosted API name Qwen3.6-Flash· Reported by Alibaba Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 35.6% |
| 13 | Qwen3.5 122B-A10B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 33.6% |
| 14 | Gemini 3 Flashthinking high· VitaBench | 32.5% |
| 15 | Qwen3.5 35B A3B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 31.9% |
| 16 | Gemini 3 Pro Previewthinking high· VitaBench | 31.5% |
| 17 | Gemini 3 Pro Previewnon-thinking low· VitaBench | 30% |
| 18 | Qwen3.5 9B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 29.8% |
| 19 | LongCat-Flash-Thinkingthinking· VitaBench | 29.3% |
| 20 | Claude Opus 4.5thinking· VitaBench | 28.5% |
| 21 | OpenAI o3thinking high· VitaBench | 26.3% |
| 22 | GPT 5.2thinking xhigh· VitaBench | 24.3% |
| 23 | DeepSeek V3.2thinking· VitaBench | 24% |
| 24 | Claude Sonnet 4.5thinking· VitaBench | 23.5% |
| 25 | Claude Opus 4.5non-thinking· VitaBench | 23.3% |
| 26 | LongCat Flash Chatnon-thinking· VitaBench | 22.8% |
| 27 | Qwen3.5 4B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 22% |
| 28 | o4 Minithinking high· VitaBench | 19.5% |
| 29 | Doubao Seed 1.8thinking high, thinking· VitaBench | 18.8% |
| 30 | DeepSeek V3.2non-thinking· VitaBench | 18.5% |
| 31 | GLM 4.7thinking· VitaBench | 18.3% |
| 32 | Claude Sonnet 4.5non-thinking· VitaBench | 17% |
| 33 | GLM 4.7non-thinking· VitaBench | 15.5% |
| 34 | Qwen 3 235B A22B Thinking 2507thinking· VitaBench | 14.5% |
| 35 | Qwen3 Maxnon-thinking· VitaBench | 14.3% |
| 36 | Doubao Seed 1.8non-thinking· VitaBench | 13.8% |
| 37 | Kimi K2 Thinkingthinking· VitaBench | 12.8% |
| 38 | Qwen3 235B A22B Instruct 2507non-thinking· VitaBench | 12.3% |
| 39 | Kimi K2 0905non-thinking· VitaBench | 11.5% |
| 40 | Qwen3 32Bthinking· VitaBench | 5.3% |
| 41 | Qwen3 32Bnon-thinking· VitaBench | 4% |
| 42 | GPT 5.2non-thinking none· VitaBench | 0.8% |
Data credits: Data by LLM Stats · BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links.