Benchmark
RISEBench - Appearance Consistency
Scores are percentages.
- Models measured
- 12
- Scores recorded
- 14
- Sources
- 1
- IQ weight
- 0.61
Score distribution
- 73.2%
- Mean
- 71.6%
- Median
- 12.9%
- Std dev
- 58.7%
- 10th percentile
- 89.3%
- 90th percentile
- 41.5%
- Min
This benchmark has a weight of 0.61 for image-generation models, measured over 10 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 | GPT Image 1.5· RISEBench data | 92.5% |
| 2 | GPT Image 2· RISEBench data | 89.3% |
| 3 | Gemini 2.5 Flash Image· RISEBench data | 86% |
| 4 | Gemini 3 Pro Image· RISEBench data | 85.5% |
| 5 | GPT Image 1· RISEBench data | 80.2% |
| 6 | BAGELw/ CoT· RISEBench data | 75.6% |
| 7 | Flux Kontext Dev· RISEBench data | 71.6% |
| 8 | GPT-Image-1 mini· RISEBench data | 71.5% |
| 9 | Qwen Image Edit2511· RISEBench data | 71% |
| 10 | Gemini 2.0 Flash Experimental· RISEBench data | 68.2% |
| 11 | Doubao Seedream 4-0· RISEBench data | 67.4% |
| 12 | Qwen Image Edit2509· RISEBench data | 66.4% |
| 13 | BAGEL· RISEBench data | 58.7% |
| 14 | Step1X-Edit· RISEBench data | 41.5% |