Theo does the math on Opus 5.5 usage limits
Half of a Claude Code subscription was off limits to Fable, and Opus 5.5 on high costs over 2x less.
Theo worked out why Opus 5.5 feels close to unlimited on a Claude Code subscription when Fable 5.1 felt rationed. His numbers, posted on X, are a roughly 4.3x increase in usable limits going from Fable 5.1 high to Opus 5.5 high, and 6.6x going from Fable 5.1 xhigh to Opus 5.5 high, which is the switch he actually made.
Fable 5.1 xhigh to Opus 5.5 high (move I made) is a 6.6x increase
Two things stack to produce that. Claude Code subscriptions are generous with usage overall, Theo says, but only half of the allowance can be spent on Fable. On top of that cap, Opus 5.5 on high is over 2x cheaper than Fable 5.1 on high. A per-model ceiling multiplied by a lower per-task cost is how you get a number that looks like a limit increase without anyone raising a limit.
The cap is the part people miss
For working engineers the useful distinction here is between price and access. Token price is the number everyone quotes, but on a subscription the binding constraint can be a model-specific share of the allowance. Theo's math says the 50% Fable restriction is doing about as much work in his 4.3x as the price difference is. Move to a model that is not capped and is cheaper per unit of work, and the same plan stretches much further, with no change to the plan itself.
This is a single user reporting his own usage, not a published allowance table, so treat the exact multipliers as Theo's arithmetic rather than a spec. His qualitative version from earlier in the week was that he had been pushing Opus 5.5 hard all day and it was barely making a dent on his limits.
The cost side, from Anthropic's numbers
Addy Osmani, summarizing the Opus 5.5 launch, said the model is 40% lower cost than Opus 5 with cache reads 60% cheaper, and that it performs at the level of Claude Fable 5.1 for most tasks. That last claim is the load bearing one for anybody considering the swap. A cheaper model only stretches your limits if it does not cost you extra turns to get the same result, and the quality parity claim is Anthropic's, not an independent measurement.
Theo's own week with the model has been busy. He also noticed a smaller change worth a mention to anyone who has been fighting the house writing style.