KVBench (Brief Caption) - History (Chinese): leaderboard
KVBench (2026) tests whether text-to-image models can draw correct senior high-school history content, such as diagrams and scientific illustrations, from Chinese prompts given as a brief caption, so the model must supply the textbook knowledge itself. Qwen2.5-VL-32B checks each image against a list of textbook-grounded constraints; the score is the percentage satisfied, higher is better.
Metric: Checklist score (%) on the history prompts in Chinese, each given as a brief caption, so the model must supply the textbook knowledge itself: KVBench knowledge-intensive text-to-image prompts from over 30 senior high-school textbooks (150 per subject and language, each with a textbook reference image); Qwen2.5-VL-32B-Instruct answers a checklist of four to six binary questions (key objects, attributes, spatial relations, reasoning outcomes) about each generated image, and the score is the mean share of satisfied items, in percent; higher is better. Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 14 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | FLUX.2-dev | 39.92 | |
| 2 | Seedream 4.0 | 35.92 | |
| 3 | GPT-Image (KVBench checkpoint unspecified) | 23.65 | |
| 4 | FLUX.1-dev | 22.83 | |
| 5 | Qwen-Image | 21.97 | |
| 6 | Z-Image | 19.97 | |
| 7 | OmniGen2 | 16.42 | #4579 |
| 8 | BAGEL-7B-MoT | 14.83 | #4619 |
| 9 | FLUX 2 Max | 10.75 | |
| 10 | Nano Banana Pro (Gemini 3 Pro Image) | 10.17 | |
| 11 | Show-o2 | 6.67 | |
| 12 | Janus-Pro | 4.83 | |
| 13 | SD3.5-Large (KVBench GGUF build, quantization unspecified) | 4.5 | |
| 14 | FLUX.1-Krea-dev | 1.17 |
Interactive version: theaggregate.ai/benchmark?slug=kvbench-brief-caption-history-chinese · How It Works · Data refreshed daily, snapshot 2026-11-21.