llm.ing

Benchmarks

Every benchmark we aggregate, what it measures, and where its numbers come from.

agentic

BenchmarkModelsUnitRun byNotes
AA Agentic Index111indexArtificial Analysis

coding

BenchmarkModelsUnitRun byNotes
AA Coding Index116indexArtificial Analysis

intelligence

BenchmarkModelsUnitRun byNotes
AA Intelligence Index228indexArtificial AnalysisComposite of independently run evals by Artificial Analysis.

preference

BenchmarkModelsUnitRun byNotes
LMArena Text137eloLMArenaOverall text category; style effects not controlled in this config.
LMArena WebDev82eloLMArenaHead-to-head web app building votes.
LMArena Vision77eloLMArenaImage-understanding battles.
Design Arena — 29 categories (crowd-voted design/code battles)
Design Arena — gamedev179eloDesign ArenaHead-to-head human votes on generated designs/code, mirrored via OpenRouter.
Design Arena — website150eloDesign ArenaHead-to-head human votes on generated designs/code, mirrored via OpenRouter.
Design Arena — codecategories140eloDesign ArenaHead-to-head human votes on generated designs/code, mirrored via OpenRouter.
Design Arena — dataviz137eloDesign ArenaHead-to-head human votes on generated designs/code, mirrored via OpenRouter.
Design Arena — 3d136eloDesign ArenaHead-to-head human votes on generated designs/code, mirrored via OpenRouter.
Design Arena — uicomponent131eloDesign ArenaHead-to-head human votes on generated designs/code, mirrored via OpenRouter.
Design Arena — svg92eloDesign ArenaHead-to-head human votes on generated designs/code, mirrored via OpenRouter.
Design Arena — asciiart86eloDesign ArenaHead-to-head human votes on generated designs/code, mirrored via OpenRouter.
Design Arena — webapps73eloDesign ArenaHead-to-head human votes on generated designs/code, mirrored via OpenRouter.
Design Arena — mobileapps51eloDesign ArenaHead-to-head human votes on generated designs/code, mirrored via OpenRouter.
Design Arena — fullstack46eloDesign ArenaHead-to-head human votes on generated designs/code, mirrored via OpenRouter.
Design Arena — androidnative44eloDesign ArenaHead-to-head human votes on generated designs/code, mirrored via OpenRouter.
Design Arena — godotgamedev35eloDesign ArenaHead-to-head human votes on generated designs/code, mirrored via OpenRouter.
Design Arena — pdf34eloDesign ArenaHead-to-head human votes on generated designs/code, mirrored via OpenRouter.
Design Arena — spreadsheet30eloDesign ArenaHead-to-head human votes on generated designs/code, mirrored via OpenRouter.
Design Arena — python-pptxslides29eloDesign ArenaHead-to-head human votes on generated designs/code, mirrored via OpenRouter.
Design Arena — agenticgamedev26eloDesign ArenaHead-to-head human votes on generated designs/code, mirrored via OpenRouter.
Design Arena — htmlslides25eloDesign ArenaHead-to-head human votes on generated designs/code, mirrored via OpenRouter.
Design Arena — codetovideo25eloDesign ArenaHead-to-head human votes on generated designs/code, mirrored via OpenRouter.
Design Arena — blender3d24eloDesign ArenaHead-to-head human votes on generated designs/code, mirrored via OpenRouter.
Design Arena — agentichtmlslides10eloDesign ArenaHead-to-head human votes on generated designs/code, mirrored via OpenRouter.
Design Arena — agenticslides(html)10eloDesign ArenaHead-to-head human votes on generated designs/code, mirrored via OpenRouter.
Design Arena — agenticslides10eloDesign ArenaHead-to-head human votes on generated designs/code, mirrored via OpenRouter.
Design Arena — agenticslides(python-pptx)10eloDesign ArenaHead-to-head human votes on generated designs/code, mirrored via OpenRouter.
Design Arena — pptxslides9eloDesign ArenaHead-to-head human votes on generated designs/code, mirrored via OpenRouter.
Design Arena — graphicdesign3eloDesign ArenaHead-to-head human votes on generated designs/code, mirrored via OpenRouter.
Design Arena — image3eloDesign ArenaHead-to-head human votes on generated designs/code, mirrored via OpenRouter.
Design Arena — logo3eloDesign ArenaHead-to-head human votes on generated designs/code, mirrored via OpenRouter.
Design Arena — imageediting1eloDesign ArenaHead-to-head human votes on generated designs/code, mirrored via OpenRouter.

speed

BenchmarkModelsUnitRun byNotes
Median output speed68tokens_per_sArtificial Analysis
Median time to first token68secondsArtificial Analysis