Fable is a beast. One-million to two-million-token runs in just 30 minutes. That’ll eat up 7 to 10 percent of your weekly allotment on a $200 subscription, in half an hour.
But it’s good, and it’s an eye-opener to know this is what it’s going to be like soon. Up to 20 dedicated agents working together to conquer that research or finish that platform build.
Some arithmetic, for anyone about to try it. The $200 plan is Anthropic’s Max 20x tier, and it carries a weekly usage limit that applies across all models and resets at a fixed time each week. The team-of-agents pattern is Anthropic’s own design, too. In February it added agent teams to Claude Code so you can “spin up multiple agents that work in parallel as a team and coordinate autonomously”, and its documentation is candid that token usage “scales with the number of active teammates”, since every teammate carries its own context window. Twenty agents is twenty meters running at once.
Now price the same half hour outside a subscription. Fable on the API is $10 per million input tokens and $50 per million output. A two-million-token run lands somewhere between $20 and $100 depending on the mix, and nobody’s mix is all input. A few of those a day and the flat fee starts to look like the bargain of the decade.
That said, I’m down to just two Claude subs and four Codex subscriptions at $200 each. Codex does more, and to a higher standard, than Claude. The Codex subscriptions are, as far as the price tells you, ChatGPT Pro, the $200-a-month tier OpenAI introduced in December 2024. Four Codex subscriptions at $200 is $800 a month before the Claude ones, which sounds profligate until you price the same tokens on a meter.
So I’m spending subscription tokens like it’s going out of business. I suspect my tap may turn off any week now, and then I’ll be paying per token via the API.
