- Anthropic’s Claude Fable 5 claimed the top spot on the Artificial Analysis Intelligence Index, surpassing rivals including GPT-5.5.
- The measured performance gain over its predecessor, Opus 4.8, is 5.7% across many benchmarks.
- Token prices have doubled: a full benchmark run now approaches $10,000, twice the cost of Opus 4.8.
- The result sharpens the question of diminishing returns at the frontier.
What Happened
Anthropic‘s latest model, Claude Fable 5, has taken the top spot on the Artificial Analysis Intelligence Index, surpassing rivals including GPT-5.5, according to The Decoder. But the gain comes at a steep price.
Compared with its predecessor, Opus 4.8, the overall measured performance leap is 5.7% across many benchmarks — while token prices have doubled.
Why It Matters
Fable 5 reclaiming the index lead matters in a race that turns over every few weeks, where Claude Opus 4.8 and OpenAI’s line have traded the crown. But a 5.7% gain for 2x the cost is the clearest sign yet that frontier improvements are getting expensive relative to what they deliver.
Technical Details
A full benchmark run on Fable 5 now approaches $10,000, twice the cost of the same run on Opus 4.8. The Artificial Analysis Intelligence Index aggregates many benchmarks into a single score, and Fable 5’s lead there sits atop GPT-5.5 — though the marginal 5.7% improvement is small next to the doubled price.
Who’s Affected
Teams running high-volume workloads feel the doubled token price directly, and many will weigh whether the 5.7% gain justifies it or whether Opus 4.8 — or a cheaper tier — remains the better value. The economics echo the long-horizon-versus-cost trade-offs surfaced by the DeepSWE benchmark.
What’s Next
The performance figures are from Artificial Analysis’s index rather than Anthropic’s own marketing, which lends them weight, but independent task-level testing will show where the 5.7% actually lands. The pricing decision tells its own story: Anthropic is betting top-tier users will pay double for the lead.
