Claude Fable 5 tops industry benchmarks with huge price gap

Anthropic’s Claude Fable 5 is the new king of AI benchmarks, crushing eight industry-specific indices from finance to medicine. Benchmarking platform Artificial Analysis built these tests using real job descriptions from the US O*NET system, weighting skills by how frequently they appear in each field. Fable 5 leads every single category, with Claude Opus 4.8 close behind and OpenAI’s GPT-5.5 snagging second in two areas. Below the top tier, the ranking shifts wildly by domain — Google, OpenAI, Anthropic’s cheaper models, and China’s GLM-5.2 all trade places depending on the task. But here’s the kicker: that top-tier quality comes with a brutal price tag. DeepSeek V4 Flash handles the same tasks for under four cents each, making you wonder if the premium is worth it. Meanwhile, Fable 5 also dominates LMArena’s real-user rankings across text, code, and agent challenges. Why it matters: The gap between elite and budget models is widening, forcing businesses to decide whether bleeding-edge accuracy matters more than cost efficiency in daily workflows.