Apple announced Monday a multi-year agreement to use Google’s Gemini models to power Siri. Anthropic announced yesterday that they have a new model tier above Opus.
Claude Fable 5 launched at $10 per million input tokens and $50 per million output, which is twice the per-token cost of Claude Opus 4.8. Anthropic explicitly calls it a “Mythos-class” model: a tier they introduced two months ago when they shipped Mythos Preview to a small group of U.S. cybersecurity partners. They made the same underlying model generally available yesterday with safeguards bolted on.
Anthropic published a Stripe customer report describing a 50-million-line Ruby codebase migration completed in a day, which Stripe estimated would have taken a team two months by hand. Anthropic also reports that Fable completed Pokémon FireRed using only raw game screenshots; earlier Claude models needed complex helper scaffolding to play at all.
The frontier model price ceiling just doubled. The Opus tier still costs $5/$25 per million tokens, the Sonnet tier still costs $3/$15, and the Haiku tier still costs $1/$5. Nothing got cheaper. Anthropic stacked a new tier on top, charged twice as much for it, and gave it a different name. If you anchored your planning on Opus economics, you are now anchored to a number that no longer represents the ceiling.
The release buries a serious issue. Fable 5 ships with classifiers that detect queries on certain topics (cybersecurity, biology, distillation attempts); on those queries, the response is automatically routed to Opus 4.8 instead and the user is informed of the switch. Anthropic estimates this happens in roughly 5% of sessions. You pay Fable rates for every query, but sometimes you get Opus work. The unit you bought is not always the unit delivered.
If you’re debating a “rent the brain” partnership, this week made the asymmetry visible. Apple announced a multi-year agreement Monday around one generation of frontier models. Twenty-four hours later, Anthropic introduced a new pricing and capability tier. Whether or not Fable proves objectively better than Gemini on independent benchmarks, the event illustrates how quickly the frontier moves relative to enterprise procurement cycles. Vendors are shipping at an accelerated cadence: weeks instead of months. Customers cannot adjust multi-year contracts that quickly. The asymmetry does not favor the buyer, and the longer the contract term, the more the asymmetry compounds.
I wrote about the hedge in Sunday’s essay. Define your own intelligence unit, anchor your procurement and your budget to it, and let the vendors keep redefining theirs. The asymmetry only works against you if you let them set the comparison.
Every company needs a Claw strategy. Do you have one?
Author’s note: This is not a sponsored post. I am the author of this article and it expresses my own opinions. I am not, nor is my company, receiving compensation for it. This work was created with the assistance of various generative AI models.