Fable 5 is Anthropic's most capable model. It sits above Opus in a new performance tier.
The shift that matters for your work: previous models were good at answering questions and drafting content one step at a time. Fable 5 can take a complex, multipart project and execute it end to end in a single session - the kind of work that used to take you five or six separate conversations spread across a day. It sustains productive work over extended periods, retains instructions across long complex tasks, and can manage parallel workstreams without losing the thread.
This is not a better content writer. Sonnet writes a perfectly good blog post. Fable 5's value is in the multistep strategic, analytical, and operational work that previously required you to be the synthesis layer - stitching together outputs from multiple sessions, manually connecting findings across data sources, or babysitting a model through a 12step process because it couldn't hold the full picture.
But there's a real catch. If you prompt Fable 5 the way you prompted previous models, you won't see the difference. Your carefully written step by step instructions don't just underperform - they actively make output worse. Anthropic's own documentation says that prompts and skills built for earlier models "are often too prescriptive for Claude Fable 5 and can degrade output quality."
The model plans better than your instructions do. So you need to stop giving it directions and start giving it destinations.
This changed two days ago. If you read anything about Fable 5 pricing before July 20, it's probably outdated.
If you're on a Pro plan: Fable 5 does not come out of your weekly limits. It's available via usage credits only - pay as you go, per token. You received a onetime $100 credit. Every Fable 5 session costs real money, so the modelrouting advice later in this guide (use Sonnet for simple tasks, save Fable 5 for complex work) is directly about your wallet.
If you're on a Max plan, Team Premium, or Enterprise Premium seat: Fable 5 is included in your weekly limits at 50% of your allocation. Combined with the fact that the bonus usage period also ended on July 20, you may feel tighter than expected. For you, the cost advice in this guide is about stretching that allocation.
The per token pricing (API and creditbased usage): $10 per million input tokens, $50 per million output tokens. That's exactly double Opus 4.8. Fable 5 at lower effort settings frequently outperforms maximum effort on previous models - so you can often get better results while spending less, if you match the effort to the task.
These are blocks you can paste into Claude Projects, into the start of a conversation, or adapt for your team's workflows. Each one solves a specific problem with how Fable 5 behaves.