Model Benchmarks

You are evaluating leading AI models as of July 9, 2026. Task: Rank the strongest publicly known models across four capability domains: 1. Coding — software engineering, debugging, multi-file refactors, agentic terminal work, and shipping production code. 2. Image generation — still images, illustration, photorealism, typography in images, and controllable style. 3. Video generation — motion quality, clip length, native audio, character consistency, and production-ready pipelines. 4. Web, animation, and physics — UI/UX implementation, CSS and canvas animation, Three.js/WebGL, interactive motion design, and believable physics in browser or real-time contexts. Requirements: - Use publicly reported benchmarks, product docs, and real-world usage where available. Do not invent benchmark numbers. - Rank the top 5 models per domain. Different domains may have different winners. - Design the chart however you want. Free reign on layout, style, density, and format (SVG, HTML/CSS, canvas, tables, whatever fits). Do not copy another model's chart design. Make it yours. - State your methodology, confidence level, and what is preview-only, deprecated, or region-locked. - Separate "best overall coding model" from "best video model" — no single model wins everything. - End with a short "best in class by domain" summary. Output structure: 1. Methodology (5–8 sentences) 2. Your chart (designed by you) 3. Domain rankings with 1–2 sentence rationale per model 4. Best-in-class summary