Claude Fable 5.2 scores 65 on Artificial Analysis, highest so far

356 points · 256 comments on HN · read original →

Points and comments are a snapshot, not live.

Claude Fable 5.2 scores 65 on Artificial Analysis Intelligence Index, setting a new high.

Artificial Analysis Intelligence Index v5.0.1 incorporates nine evaluations. Claude Fable 5.2 (max) scores 65, the highest to date. The model uses significantly more output tokens per task (274k) than Fable 5.1 (118k), driving higher costs. No publicly available API is listed. The page shows pricing, context window, and performance metrics including output speed and time per task.

What commenters are saying

Commenters note the rapid release cadence, with one likening it to the CPU clock speed race that eventually slowed. The high cost per task is a major discussion point; several note that the "max" setting is expensive and the medium setting is far more efficient. Some defend Anthropic, saying most plan users don't exhaust usage. Google's position is debated: some find Argon useful for specific tasks like debugging and writing, while others say Google botched its release and lags in code generation.