Claude Fable 5.1 is impressive. You don't need it.

Anthropic has released Claude Fable 5.1, which it calls "the world's most advanced models for coding and knowledge work". The plural is because it arrived alongside a twin, Claude Mythos 5.1. They're the same model. Mythos has looser safety controls and is only open to vetted organisations doing cyber security and life sciences work. Fable is the one everyone else can use.

We think these frontier models are cool. We also think the hype around each release is nonsense. Most people don't need a new model. They need to know how to use the one they've already got.

What actually changed

The improvements are real. Anthropic says Fable 5.1 is better at coding, at knowledge work and at tasks that run for a long time. It keeps Fable 5's price, and Anthropic estimates it will cost about 25% less for typical workloads. Its safety checks also get in the way less often when the request is harmless.

The headline benchmark is a test called Terminal-Bench-Science, where Fable 5.1 scores 52.6% against Fable 5's 24.7%. That's a big jump. It's also a test of scientific computing tasks run in a command line. If your week is quotes, emails, rotas and invoices, it tells you nothing about whether the model will help you.

Why a new model rarely fixes anything

When AI disappoints someone at work, the model is almost never the problem. The usual causes are much more ordinary:

• No context. It doesn't know your business, your prices, your customers or how you write, so it guesses.
• Vague asks. "Write a proposal" gets you a generic proposal. A proposal for a named client, with the brief, the budget and an example you liked, gets you something usable.
• Taking the first answer. People take the first answer and give up, when the second or third round is where it gets good.
• Asking it to agree. AI is very keen to tell you your idea is great. We built a tool to deal with exactly that.

A smarter model fixes none of these. Given no context, a cleverer model still guesses, just more fluently.

What to do instead

If you'd been planning to try the new release, spend that time on this:

1. Set up a project. Claude and ChatGPT both let you keep documents and standing instructions in one place, so you stop starting from scratch. Put in the things a new member of staff would need: what you sell, what you charge, who your customers are, and a few examples of writing you're happy with.
2. Write proper instructions. Say who it's working for, what good looks like and what to avoid. A few short paragraphs is plenty.
3. Give it real work. Pick one job you do every week and do it with AI for a month. You'll learn more from that than from any launch post.
4. Check its work. Read what it gives you as you would a capable new starter's: useful, quick, and occasionally confidently wrong.

If you're still deciding which assistant to use, we've written about why Claude is worth the switch from ChatGPT. The honest answer is that any of the main ones will do far more for you than you're getting now, once it's set up properly.

So who should care about Fable 5.1?

The improvements are aimed at people building software with AI and people running long research or technical tasks. Businesses paying for AI by usage get lower running costs. For everyone else, it's good news that will quietly make the tools you already use a bit better, without you having to do anything.

The next release will be impressive too, and learning to use the one you have will still do more for your business than waiting for it.

If there's a job in your business you think AI should be able to do and you can't get it working, we build tools like that.