← Latest AI News

November 10, 2026 · Daily brief

Claude just got 20% cheaper and 30% faster at hard tasks

Sovereignty angle
Cheaper and faster sounds like a deal, but the model still sees every prompt you send. Do not paste the customer database or the full contract. Hand it the summary, the question, the draft. Keep the original with you.

Anthropic cut Fable 5.2 pricing to eight dollars input and forty output per million tokens. The model runs up to 30% faster than the previous version.

What changed

Anthropic released Claude Fable 5.2 on November 9. Input tokens dropped from ten dollars to eight, output from fifty to forty per million. That is a 20% cut on both sides. The model also generates answers up to 30% faster than Fable 5.1, released in September.

Fable 5.2 is the public version. Mythos 5.2 is the restricted sibling for vetted security and research teams. Same engine, different safety rules.

What you can do with it now

Give Fable the tasks where speed and cost both matter. Turn a messy client email into a clear three-line reply. Summarize a ten-page proposal into bullet points your team can discuss. Draft the outline for a report you will finish yourself.

Fable now costs eight dollars per million input tokens. At that rate, you can paste a hundred pages of text for under two dollars. If the task used to take your team an hour, and Fable gives you a usable draft in ten seconds, the math is simple.

What to keep in your own hands

Fable is a hosted model. Anthropic sees the prompts, stores some data for trust and safety, and controls the infrastructure. Do not send confidential client files, employee records, or anything that would cost you a customer if it leaked.

Send the question, not the database. Send the template, not the signed contract. Review every draft before you hit send. The model writes fast, but it does not know what your client actually needs or what promise you made last month. You do.