KVBench (Detailed Caption) - Math (English): leaderboard

KVBench (2026) tests whether text-to-image models can draw correct senior high-school math content, such as diagrams and scientific illustrations, from English prompts given as a detailed caption that spells out the required visual content. 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 math prompts in English, each given as a detailed caption that spells out the required visual content: 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. 16 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 2.5 Flash Image67.96
2GPT-Image (KVBench checkpoint unspecified)66.21
3Seedream 4.066.05
4FLUX.2-dev58.56
5Nano Banana Pro (Gemini 3 Pro Image)53.11
6FLUX 2 Max52.91
7FLUX.1-Krea-dev44.48
8SD3.5-Large (KVBench GGUF build, quantization unspecified)39.97
9HiDream (KVBench checkpoint unspecified)35.04
10FLUX.1-dev30.17
11BAGEL-7B-MoT25.59#4619
12Qwen-Image23.13
13Z-Image15.32
14OmniGen26.6#4579
15Show-o25.63

Interactive version: theaggregate.ai/benchmark?slug=kvbench-detailed-caption-math-english · How It Works · Data refreshed daily, snapshot 2026-11-21.