Anthropic released two new AI models this week, Claude Fable 5.1 and Claude Mythos 5.1, and buried in the announcement was a detail that matters far more to most UK small businesses than any benchmark score: repeated use of the same context now costs a quarter of what it did before. If your business touches AI tools that quote Claude’s pricing, whether directly through a developer or indirectly through a SaaS product built on top of it, that’s worth understanding.

Fable 5.1 is the generally available model, described by Anthropic as its most capable yet for coding and knowledge work, while Mythos 5.1 is a more capable sibling restricted to vetted cybersecurity and life-sciences organisations. Both support a million tokens of context, meaning they can hold an entire codebase, a long contract, or a full year of correspondence in working memory at once. For most SME purposes, though, the pricing change is the part that actually shows up on an invoice.

Why the “cache” detail matters more than it sounds

AI tools that repeatedly reference the same background information, your company’s product catalogue, a support knowledge base, a codebase, can reuse that stored context instead of resending it every time. That’s called a cache hit, and Anthropic cut the cost of one from $1.00 to $0.25 per million tokens for Fable 5.1. That’s a 75% drop on exactly the kind of repeated, high-volume use a customer service bot or internal coding assistant relies on.

The sticker price for fresh requests hasn’t moved, still $10 and $50 per million tokens depending on the model. So this isn’t a headline price cut you’d notice on a comparison chart. It’s a change that mainly benefits tools doing lots of repeat lookups against the same material, which describes a large share of the AI features quietly built into everyday business software.

What to actually do with this

Ask your AI-tool suppliers whether they’ve passed the saving on. If you pay a monthly fee for an AI feature inside your CRM, helpdesk, or accounting software, and that feature is built on Anthropic’s models, the vendor’s costs may have just dropped. Whether that shows up as a lower bill for you depends entirely on their pricing structure, and it’s a fair question to ask directly rather than assume.

If you’re building anything in-house, revisit the numbers. Businesses working with CoolCoding or an internal developer on a custom AI feature, a document assistant, a support tool, an internal search system, should have their build costed against Fable 5.1’s new cache pricing. A project that looked marginal on cost six months ago may now clear the bar, particularly anything that repeatedly references the same documents or codebase.

Don’t assume newer automatically means better for your task. Fable 5.1 outperforms its predecessor on coding and agentic benchmarks, but a frontier model still isn’t always the right tool for a narrow, repetitive job like invoice categorisation or ticket triage. BuildApps works with UK SMEs specifically on matching the model to the job rather than defaulting to whatever launched most recently.

Treat AI pricing as a moving target, not a fixed cost. This is at least the third meaningful pricing shift from a major AI lab in the past two months. If a supplier’s contract or your own budgeting assumes today’s AI costs are stable for the next year, build in room for them to move again, in either direction.

The takeaway

The headline from this launch is a smarter, more capable model. The number that actually affects your business is a quiet 75% cut to repeat-use pricing that most coverage won’t mention. If AI tooling is part of your cost base, whether bought in or built in-house, it’s worth five minutes to check whether that saving has reached you yet.