Claude Fable 5.1 tops Intelligence Index, costs 20% more per task

Anthropic’s Claude Fable 5.1 just claimed the top spot on the Artificial Analysis Intelligence Index with a score of 66, beating out more than 192 competing models. That’s the good news. The less convenient news: it costs roughly 20% more per task than its predecessor, despite a massive 75% cut to cache read prices. The model, released on September 1, 2026, averages $3.69 per task on the Intelligence Index. Fable 5, which launched in June, ran between $3.08 and $3.25 per task. The pricing paradox On paper, Fable 5.1’s pricing looks identical to its predecessor. Input tokens still cost $10 per million. Output tokens still cost $50 per million. Those numbers haven’t budged since Fable 5 debuted on June 9. Where Anthropic did cut prices is on cache reads, slashing them from $1 per million tokens down to $0.25 per million tokens. That 75% reduction is genuinely meaningful for users who repeatedly query large datasets or maintain persistent context across sessions. But the per-task cost increase tells a different story. Fable 5.1’s enhanced capabilities in long-horizon agentic coding and multi-step research mean the model tends to generate significantly more output when tackling complex...