Benchmark index
Every benchmark we run. A new benchmark opens roughly every two weeks. Live benchmarks accumulate judgements on a fixed prompt set; each release is versioned and permanently citable.
Image
1 live · 4 coming soonGeneral image
Leaderboards: overall preference, coherence, prompt alignment
Text-to-Image · v1 · 2026-07-23Text-to-Image
Models
Prompts
Judgements
Text renderingComing soon
Text accuracy, legibility, layout
Text-to-Image
Variance (Mode Collapse)Coming soon
Output diversity across seeds
Text-to-Image
Image editingComing soon
Instruction adherence, edit quality
Image-to-image
Facial preservationComing soon
Identity retention
Image-to-image
World Models
4 liveCamera movement
Leaderboards: movement match
Camera Movement · v1 · 2026-09-03Camera Movement
Models
Prompts
Judgements
Relative camera movement
Leaderboards: prompt alignment
Relative Camera Movement · v1 · 2026-09-30Relative Camera Movement
Models
Prompts
Judgements
World Model Physics
Leaderboards: blind, with-reference
Physics · v1 · 2026-09-30Physics
Models
Prompts
Judgements
World Models Realism
Leaderboards: photorealism, visual integrity
Realism · v1 · 2026-10-06Realism
Models
Prompts
Judgements
Video
2 coming soonGeneral image-to-videoComing soon
Preference, temporal coherence
Image-to-Video
General videoComing soon
Preference, temporal coherence
Text-to-Video
Image-to-3D
1 liveSVG
1 live · 1 coming soonSVG generation
Leaderboards: overall preference, coherence, prompt alignment
SVG · v1 · 2026-08-20SVG
Models
Prompts
Judgements
Animated SVGComing soon
Motion fidelity
SVG
LLM-to-3D
1 liveLive benchmarks update continuously; frozen rounds remain available as citable snapshots from the benchmark page.
Want a benchmark that is not listed? Suggest a benchmark.
We use crowd intelligence to deliver real, diverse human feedback at scale for model evaluation and post-training.
Benchmarks
Rapidata, benchmark.ai © 2026