Why it matters
The pitch is cost per task, not sticker price. On FrontierCode 2.0, Anthropic says Fable 5.2 at High effort scores 45.3%, a hair under the 45.5% best from Opus 5.5, at about half the spend per task. Vendor-run numbers; test them on your own workload.
Tokenmaxxing read
The higher-tier model can be the cheaper job, yet Anthropic still says to start most work on Opus 5.5 and move up to Fable or down to Sonnet 5.5 or Haiku 5.5 as scope demands. It also casts Fable as an advisor smaller models consult. That is a vendor-written routing table.
Source takeaway
Effort is the dial to watch. Anthropic reports Fable 5.2 at Medium effort, the Claude Code default, matching Opus 5.5's top WANDR score at roughly 40% of its per-task cost, and at Low effort matching its top Humanity's Last Exam score for under a tenth.


