Glossary
Plain names for the fifty slice ids in or_slices.json. Each sentence says what that cut ranks and which direction is better. A missing score stays blank. Numbers live on the benchmark board, not here.
Artificial Analysis
- aa-intelligence
- The aa-intelligence slice ranks models on the Artificial Analysis intelligence index, and a higher index is better.
- aa-coding
- The aa-coding slice ranks models on the Artificial Analysis coding index, and a higher index is better.
- aa-agentic
- The aa-agentic slice ranks models on the Artificial Analysis agentic index, and a higher index is better.
Design Arena models
- da-models-3d
- The da-models-3d slice ranks models by Design Arena Elo for 3D generation, and a higher Elo is better.
- da-models-3d-speed
- The da-models-3d-speed slice ranks the same 3D runs by average generation time in milliseconds, and a lower time is better.
- da-models-asciiart
- The da-models-asciiart slice ranks models by Design Arena Elo for ASCII art, and a higher Elo is better.
- da-models-asciiart-speed
- The da-models-asciiart-speed slice ranks those ASCII-art runs by average generation time in milliseconds, and a lower time is better.
- da-models-codecategories
- The da-models-codecategories slice ranks models by Design Arena Elo across code categories, and a higher Elo is better.
- da-models-codecategories-speed
- The da-models-codecategories-speed slice ranks those code-category runs by average generation time in milliseconds, and a lower time is better.
- da-models-dataviz
- The da-models-dataviz slice ranks models by Design Arena Elo for data visualizations, and a higher Elo is better.
- da-models-dataviz-speed
- The da-models-dataviz-speed slice ranks those visualization runs by average generation time in milliseconds, and a lower time is better.
- da-models-gamedev
- The da-models-gamedev slice ranks models by Design Arena Elo for game development, and a higher Elo is better.
- da-models-gamedev-speed
- The da-models-gamedev-speed slice ranks those game-development runs by average generation time in milliseconds, and a lower time is better.
- da-models-svg
- The da-models-svg slice ranks models by Design Arena Elo for SVG drawings, and a higher Elo is better.
- da-models-svg-speed
- The da-models-svg-speed slice ranks those SVG runs by average generation time in milliseconds, and a lower time is better.
- da-models-uicomponent
- The da-models-uicomponent slice ranks models by Design Arena Elo for UI components, and a higher Elo is better.
- da-models-uicomponent-speed
- The da-models-uicomponent-speed slice ranks those UI-component runs by average generation time in milliseconds, and a lower time is better.
- da-models-website
- The da-models-website slice ranks models by Design Arena Elo for websites, and a higher Elo is better.
- da-models-website-speed
- The da-models-website-speed slice ranks those website runs by average generation time in milliseconds, and a lower time is better.
Design Arena agents
- da-agents-agenticgamedev
- The da-agents-agenticgamedev slice ranks agent runs by Design Arena Elo in the agentic game-dev category, and a higher Elo is better.
- da-agents-agentichtmlslides
- The da-agents-agentichtmlslides slice ranks agent runs by Design Arena Elo in the agentic HTML slides category, and a higher Elo is better.
- da-agents-agenticslides
- The da-agents-agenticslides slice ranks agent runs by Design Arena Elo in the agentic slides category, and a higher Elo is better.
- da-agents-agenticslides(html)
- The da-agents-agenticslides(html) slice ranks agent runs by Design Arena Elo in the HTML agentic-slides category, and a higher Elo is better.
- da-agents-agenticslides(python-pptx)
- The da-agents-agenticslides(python-pptx) slice ranks agent runs by Design Arena Elo in the python-pptx agentic-slides category, and a higher Elo is better.
- da-agents-androidnative
- The da-agents-androidnative slice ranks agent runs by Design Arena Elo for native Android apps, and a higher Elo is better.
- da-agents-blender3d
- The da-agents-blender3d slice ranks agent runs by Design Arena Elo for Blender 3D work, and a higher Elo is better.
- da-agents-fullstack
- The da-agents-fullstack slice ranks agent runs by Design Arena Elo for full-stack apps, and a higher Elo is better.
- da-agents-godotgamedev
- The da-agents-godotgamedev slice ranks agent runs by Design Arena Elo for Godot games, and a higher Elo is better.
- da-agents-htmlslides
- The da-agents-htmlslides slice ranks agent runs by Design Arena Elo for HTML slide decks, and a higher Elo is better.
- da-agents-mobileapps
- The da-agents-mobileapps slice ranks agent runs by Design Arena Elo for mobile apps, and a higher Elo is better.
- da-agents-pptxslides
- The da-agents-pptxslides slice ranks agent runs by Design Arena Elo for PPTX slide decks, and a higher Elo is better.
- da-agents-python-pptxslides
- The da-agents-python-pptxslides slice ranks agent runs by Design Arena Elo for slide decks built with python-pptx, and a higher Elo is better.
- da-agents-spreadsheet
- The da-agents-spreadsheet slice ranks agent runs by Design Arena Elo for spreadsheets, and a higher Elo is better.
- da-agents-webapps
- The da-agents-webapps slice ranks agent runs by Design Arena Elo for web apps, and a higher Elo is better.
OpenRouter
- or-gpqa_diamond
- The or-gpqa_diamond slice ranks models by accuracy on GPQA Diamond, and a higher accuracy is better.
- or-gpqa_diamond-cost
- The or-gpqa_diamond-cost slice ranks models by average cost per GPQA Diamond task, and a lower cost is better.
- or-tau_bench_verified_airline
- The or-tau_bench_verified_airline slice ranks models by accuracy on the verified airline tau-bench tasks, and a higher accuracy is better.
- or-tau_bench_verified_airline-cost
- The or-tau_bench_verified_airline-cost slice ranks models by average cost per verified airline tau-bench task, and a lower cost is better.
- or-search_browsecomp
- The or-search_browsecomp slice ranks search runs by accuracy on BrowseComp, and a higher accuracy is better.
- or-search_browsecomp-cost
- The or-search_browsecomp-cost slice ranks those BrowseComp runs by average cost per task, and a lower cost is better.
- or-search_browsecomp-latency
- The or-search_browsecomp-latency slice ranks those BrowseComp runs by average latency per task in milliseconds, and a lower latency is better.
- or-search_hle
- The or-search_hle slice ranks search runs by accuracy on Humanity's Last Exam, and a higher accuracy is better.
- or-search_hle-cost
- The or-search_hle-cost slice ranks those exam runs by average cost per task, and a lower cost is better.
- or-search_hle-latency
- The or-search_hle-latency slice ranks those exam runs by average latency per task in milliseconds, and a lower latency is better.
- or-search_dsqa
- The or-search_dsqa slice ranks search runs by accuracy on DeepSearchQA, and a higher accuracy is better.
- or-search_dsqa-cost
- The or-search_dsqa-cost slice ranks those DeepSearchQA runs by average cost per task, and a lower cost is better.
- or-search_dsqa-latency
- The or-search_dsqa-latency slice ranks those DeepSearchQA runs by average latency per task in milliseconds, and a lower latency is better.
- or-search_widesearch
- The or-search_widesearch slice ranks search runs by item-level F1 on WideSearch, and a higher F1 is better.
- or-search_widesearch-cost
- The or-search_widesearch-cost slice ranks those WideSearch runs by average cost per task, and a lower cost is better.
- or-search_widesearch-latency
- The or-search_widesearch-latency slice ranks those WideSearch runs by average latency per task in milliseconds, and a lower latency is better.