How to install Anthropic launches Claude Fable 5.1 and says it’s up

What happened

I'll write the guide from the given facts only, then count words so it lands in the 750–900 range.The first draft ran long. I'll cut it to 750–900 words and keep the required sections.Anthropic has launched Claude Fable 5.1, and The Verge reports that the company says this model is up to 45 percent cheaper for agentic work. Mythos 5.1 shipped in the same announcement, so the newest models are a pair rather than a single name swap.

The company is answering customer criticism on price, data retention, and overzealous safeguards. Anthropic claims Claude Fable 5.1 offers stronger performance than Fable 5, costs around 25 percent less typically, and costs up to 45 percent less for complex agentic tasks thanks to reduced pricing.

Claude Fable 5.1 and Mythos 5.1 are the newest AI models named in the report. The cost and performance claims sit on Fable 5.1: stronger than Fable 5, around 25 percent less in typical use, and up to 45 percent less on complex agentic tasks because of reduced pricing. Mythos 5.1 is part of the same launch and is not given those figures. Price, data retention, and overzealous safeguards are the three customer criticisms Anthropic says this release addresses.

If you already call Fable 5, you now have a named successor with two different savings numbers. Typical jobs are the around 25 percent case. Complex agentic jobs are the up to 45 percent case. Stronger performance than Fable 5 is claimed only for Fable 5.1.

If price was your complaint, those percentages are the company's answer, and your existing jobs are the only way to check them. The report does not break those percentages down further.

How it works

If data retention was your complaint, treat Fable 5.1 and Mythos 5.1 as in scope and inspect the retention controls on your account. The report names the issue. It does not describe the new rule.

If overzealous safeguards were your complaint, re-run the Fable 5 prompts that used to be refused, with the text unchanged, on Fable 5.1 and on Mythos 5.1. Do not fold both models into one upgrade flag. Only Fable 5.1 carries the published cost and performance claims.

List every call, config value, and router rule that still names Fable 5. That list is the upgrade surface.

Point those entries at Fable 5.1 in the same places you already select a model. If you also need Mythos 5.1, add it as its own choice. Do not rename Fable 5 to Mythos 5.1.

Why it matters

Keep Fable 5 available. You cannot test the stronger-performance claim if you remove the baseline before you run paired jobs.

Take typical jobs from work you already bill and use them to check the around 25 percent claim. Take complex agentic jobs from work you already bill and use them to check the up to 45 percent claim. Run each set on Fable 5, then on Fable 5.1, and record cost. The upgrade is not done until both comparisons exist.

While those jobs run, watch refusals. A cheaper model that still blocks the same agentic tool calls does not close the safeguards complaint. Before you send customer data, confirm retention settings for Fable 5.1 and Mythos 5.1 against the policy you already promised users.

Only after the two cost comparisons, the Fable 5 performance comparison, the safeguard check, and the retention check should Fable 5.1 become the default for agentic work. Default Mythos 5.1 only on flows you chose for that model.

Who is affected

Forty-five percent is not the number for every invoice. Around 25 percent less is the typical figure. Up to 45 percent less applies to complex agentic tasks. If your workload is not that class of work, the higher figure is the wrong target.

Mythos 5.1 does not inherit the Fable 5.1 cost story or the stronger-performance claim. Those claims are Fable 5.1 versus Fable 5.

Reduced pricing is the stated reason for the savings. If the bill does not move, check that the run is not still calling Fable 5, that the jobs match the typical or complex agentic cases, and that another line item is not dominating the invoice.

An upgrade that only changes model names and skips retention is incomplete relative to the criticisms Anthropic says it is answering. Safeguards are named without a published pass-rate, so existing tools and prompts may still be refused. Test the same tool-using flows you already run.

What to watch next

The report does not give a date, a package version, a model ID string, or a regional availability list. Do not invent those. Use the names Fable 5, Fable 5.1, and Mythos 5.1 as reported, and take install mechanics from Anthropic's own interface.

Keep typical-job invoices against the around 25 percent claim and complex agentic invoices against the up to 45 percent claim on separate lines. Mixing the two figures will hide whether either one is holding.

Watch whether Fable 5.1 is stronger than Fable 5 on the agentic tasks you already run. Watch retention language for both new models in your account, because the launch names that criticism and the summary does not spell out the new rule. Watch the same prompts that used to fail; if refusals stay high, the safeguards complaint is still open even if the bill is lower.

Keep Mythos 5.1 on its own ledger. It shipped with Fable 5.1 and does not carry the published 25 percent and 45 percent figures.