Update - 13 June: Fable 5 is temporarily unavailable. On 12 June the US government issued an export control directive requiring Anthropic to suspend access to Fable 5 (and Mythos 5) for all customers, so it's currently switched off across Chat Thing along with everywhere else. Every other model, Opus 4.8 included, is unaffected. Anthropic believe it's a misunderstanding and are working to restore access; we'll switch it back on the moment they do.
Anthropic just shipped Claude Fable 5, their most capable model yet, and it's live on Chat Thing today. It's a proper step up, and it arrives with a wave of other new models and a rebuilt model picker to help you choose between them.
Claude Fable 5: the new most-capable model
Fable 5 is the most capable model Anthropic have released. It's built for the demanding stuff, long multi-step agentic work and proper hard reasoning, the kind of task that used to fall over halfway through. It's got a 1 million token context window, so it can hold a huge amount of your data in mind at once, and it has a good think before it answers.
If you've got a bot where getting the answer exactly right matters more than getting it cheap, this is the one to point it at.
One thing we'd rather you heard from us, though. Fable is expensive, and it gets through tokens quickly. It costs roughly twice as much per message as Claude Opus 4.8 (our next model down) and many times more than the everyday models. That power is brilliant when you genuinely need it, but it'll chew through your token allowance much faster, so it's worth keeping Fable for the jobs that really call for it and using a lighter model for routine chat. The new picker shows you exactly how each model is priced, so there shouldn't be any surprises.
The rest of the new line-up
Fable 5 is the headline, but it's not arriving alone. We've added the latest from across the board, all live now:
- Claude Opus 4.8 (plus a Fast variant) - Anthropic's most capable Opus model, highly autonomous and great at long agentic runs and holding context across a whole task
- GPT-5.5 and GPT-5.5 Pro from OpenAI, both with a 1M-plus context
- Gemini 3.5 Flash from Google - fast, cheap, 1M context, a lovely default for high-volume bots
- Grok 4.3 from xAI, with a 1M context and keen pricing
- Mistral Medium 3.5, a strong, well-priced all-rounder
- Kimi K2.6, GLM 5.1, and DeepSeek V4 Pro - the last of those is great value if cost is a concern, with a 1M context for a fraction of the price
A model picker that actually helps you choose
The old selector was a single dropdown, a long flat list of names. Fine with a handful of models, but we're well past a handful now, and a flat list tells you nothing about which one to pick. You couldn't see what a model cost, how big its context was, or whether it could handle images, without choosing it and finding out.
So we rebuilt it. Open your agent settings, click the model, and a proper picker opens in place. You can browse by provider down the side (with a live count of how many each one has), search by name, and filter by what you actually need, whether that's Vision, Tools or Reasoning. Every card shows the input and output cost, the context size and a simple Low / Medium / High badge, so you can weigh a cheap, fast model against a premium one before you commit. It's also where you'll see at a glance just how pricey Fable is next to the rest. And we float our recommended models to the top, so there's always a sensible default if you'd rather not think about it.
Who gets it
Model selection is available on our Standard, Pro and Enterprise plans. If you're on Free or Basic your bots run on our default model and you won't see the picker, so it's worth a look at the plans if you'd like the freedom to choose.
That's the lot: a stack of new models with the best Anthropic have ever made sitting at the top, and a picker built to help you choose wisely (and keep an eye on the bill). Open an eligible agent, click the model, and take Fable 5 for a spin.
