Claude Opus 5 leads on agentic work — and undercuts Fable 5 on cost

Short summary

Claude Opus 5 tops Artificial Analysis benchmarks for agentic knowledge work at 61 on the Intelligence Index, tying Claude Fable 5 and beating GPT-5.6 Sol, while costing 26% less per task at $2.03 vs $2.75. It leads on agentic Elo scores and ties for first on coding, but its 50% hallucination rate makes it unsuitable for factual precision tasks where Fable 5 still holds the edge.

• •Opus 5 leads agentic benchmarks at 1861 Elo on GDPval-AA v2, 100+ points ahead of Fable 5 and GPT-5.6 Sol
• •Cost per task is $2.03 at max effort, 26% cheaper than Fable 5's $2.75
• •50% hallucination rate is a hard limit for factual knowledge tasks — Fable 5 still wins there

Generated with AI, which can make mistakes.