X Discussion: Theo Analyzed 1.1 Million Requests and Says the $200 Claude Ultra Plan Buys About $15,000 of Opus 5.5 a Month but Only About $3,000 of Fable 5.2, Arguing Work Done per Week Is the Number That Matters

Core Background and Industry Pain Points

After Claude Fable 5.2 shipped on November 9, heavy users cared less about benchmarks than about how long subscription limits would last. T3 Code CEO Theo (@theo) says he logs every request he makes in Claude Code, and on November 10 (09:09 SGT) he used that log to compare how fast the two models drain an Ultra subscription. His first post said Fable 5.5 by mistake, which he corrected to 5.2.

Architecture Highlights and Underlying Mechanisms

Theo split the question three ways: dollars versus percent of the limit, tokens versus percent, and the Fable usage cap. His conclusion is that Fable burns limit faster per dollar than Opus. Anthropic's Help Center says that during launch week, Pro plans and standard Team seats can use Fable for up to 50% of weekly usage limits, while Max and Ultra users get one to three Fable limit resets, and a reset returns only the usage spent on Fable models.

Authoritative Benchmarks and Measured Results

Anthropic has not published subscription burn rates, so Theo's numbers are a personal sample. The official figures: on FrontierCode 2.0, Fable 5.2 at high effort scores 45.3%, close to Opus 5.5 at max effort (45.5%), at about half the cost per task. That is part of why the gap narrows when he counts turns completed per week instead of dollars.

Developer Adoption and Getting Started

Anthropic's own model guide also recommends starting with Opus 5.5 and moving to Fable 5.2 for harder work. Replies suggested using Fable only as an orchestrator while Opus or Sonnet runs subtasks, to save Fable usage. During launch week, save Fable resets for the hardest long-running tasks, and when buying plans for a team, compare the work completed per week, not just the API-equivalent dollar value.