image-bench-ai
August 26, 2026 · View on GitHub
Reproducibility artifacts for ImageBench — a public leaderboard for text-to-image models. This repo holds the frozen prompt suites and judge questions used to score models; the results, the generated images, and the interactive leaderboard live on the site.
Latest results: imagebench.ai — 50+ models scored across every category, every generated image published in the gallery, nothing cherry-picked.
What ImageBench is
- A fixed benchmark: same 192 generation prompts, same specialized judge questions, same scoring protocol for every model.
- A capability probe, not a beauty contest — the categories target things text-to-image models actually get wrong (text rendering, counting, negation, hands, physics, reflections, world knowledge).
- Automated grading by vision-language model (VLM) judges. Every image gets a binary PASS/FAIL verdict on a concrete, test-specific question.
- Everything is published: every generated image, every judge verdict, every aggregated score. No selection, no retries, no post-hoc tuning per model.
The headline Overall score is a blend:
Overall = 0.5 × Capability + 0.5 × EPS
where Capability is the VLM PASS rate over the 192 tests and EPS (Estimated Preference Score) is an aesthetic axis from the HPSv3 human-preference reward model. Details: imagebench.ai/blog/benchmark-v1-methodology.
The versions
ImageBench is a single benchmark that has evolved. All versions share the same 64 tests and 192 generation prompts — what changes is how the judge grades them.
| Version | Judge | What changed | Read more |
|---|---|---|---|
| V1 | Multi-VLM router (Qwen3-VL, Qwen3.5-122B, Gemma4-26B, Qwen3.6-27B) per category | Original release. Generic per-category question template with a {prompt_excerpt} placeholder — the judge sees the generation prompt and interprets it. | imagebench-v1/ · methodology |
| V1.1 | Gemini 3.1 Pro (single frontier judge) | Specialized, standalone question per prompt variant (question_1/2/3). The judge never sees the generation prompt — the question bakes in the ground truth. Stricter, more consistent grading; scores drop ~22 points but the ranking largely holds. | imagebench-v1.1/ · blog post |
| V1.2 | V1.1 + Qwen3.5-122B on Hands | Gemini over-flagged hands (F1 65% vs Qwen3.5-122B's 80% on human-labeled hands). V1.2 routes only the 12 Hands tests to Qwen3.5-122B; the other 180 tests stay with Gemini. Human-realism scores rise ~18 points; ranking barely moves. | Same prompts as V1.1 · blog post |
V1.2 does not add a new prompt file — it reuses imagebench-v1.1/prompts.csv and only changes
the per-category judge routing. Prompts and judge questions are frozen across versions; only
the grader varies.
Repo layout
imagebench-v1/
challenges.csv # V1 frozen challenge suite (192 rows, one per prompt variant)
imagebench-v1.1/
prompts.csv # V1.1 benchmark definition (64 tests × 3 variants, one row per test)
categories.csv # category / subcategory / difficulty descriptions
METHODOLOGY.md # V1.1 spec, schema, and how a judge sees a prompt
imagebench-v1/challenges.csv and imagebench-v1.1/prompts.csv are two shapes of the same
prompts:
- V1
challenges.csvis flattened — one row per (test × variant), so 192 rows. Each row carries the resolvedvision_questionfor that specific variant using the V1 generic template, plus the V1 router's assigned VLM. - V1.1
prompts.csvis wide — one row per test (64 rows), withprompt_1/2/3and the matching specializedquestion_1/2/3side by side. Thevision_question_templatecolumn keeps the V1 generic form as a fallback for any variant whose specialized question is blank.
If you're reproducing V1 exactly, use the V1 file. If you're running the current benchmark or adding a new specialized question, use the V1.1 file.
Reproducing a run
- Generate 192 images from either
imagebench-v1/challenges.csv(one image per row) orimagebench-v1.1/prompts.csv(three images per row — one perprompt_N). - Ask a VLM judge the corresponding
vision_question(V1) orquestion_N(V1.1) about each image, with the system-prompt convention that the judge ends its response withVERDICT: PASSorVERDICT: FAIL. - Aggregate: score = PASS count / images evaluated, reported overall and per category / subcategory / difficulty.
- Compare against the published runs at imagebench.ai. Every image and every verdict for the 50+ scored models is downloadable via the gallery pages.
For the aesthetic EPS axis, run the images through HPSv3
and blend as Overall = 0.5 × Capability + 0.5 × EPS.
Attribution
If you use these prompts or reproduce ImageBench results, please cite imagebench.ai and link back.
Links
- Leaderboard & full results: https://imagebench.ai
- Gallery (every image, every model): https://imagebench.ai/gallery
- V1 methodology: https://imagebench.ai/methodology-v1
- V1.1 change (Gemini judge): https://imagebench.ai/blog/benchmark-v1-1-gemini-judge
- V1.2 change (Hands → Qwen): https://imagebench.ai/blog/benchmark-v1-2-hands-qwen
- Aesthetic axis (EPS): https://imagebench.ai/blog/estimated-preference-score