Live arena rankings across text, code, vision, image, and video, translated for the marketer who needs to know which model to put on which job by Monday morning.
Lead with Anthropic's Claude Opus 4.6/4.7 thinking. Top-3 in Text, Document, and Vision. Stop A/B-ing models for prose work.
WebDev and Image-to-WebDev are an Anthropic monopoly. Use Opus 4.7 for mockup-to-app; fall back to Kimi K2.6 on cost.
GPT-Image-2 (medium) is the only image model topping both Text-to-Image and Image-Edit. Use Gemini Nano-Banana-Pro for fast iteration.
Bytedance Dreamina Seedance 2.0 is #1 in all three video arenas. Veo 3.1 is the credible Google fallback for brand-safe accounts.