Claude Fable 5.2 is faster and cheaper. What it changes for your business

Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 on 9 November 2026. Fable 5.2 is the model most businesses can actually use. It generates output up to 30 percent faster than Fable 5.1 and costs 20 percent less: 8 US dollars per million input tokens and 40 per million output tokens, down from 10 and 50. For a small business that uses AI every day for coding, drafting and research, that combination matters more than any single benchmark score.

What changed in Fable 5.2

The price falls from 10 to 8 dollars per million input tokens and from 50 to 40 dollars per million output tokens. Output generation is up to 30 percent faster. The context window holds one million tokens, and the model ID is claude-fable-5-2.

The benchmarks come from Anthropic itself, so read them with care. On FrontierCode 2.0, a test for agentic coding, Anthropic reports Fable 5.2 at 49.1 percent, against 32.4 percent for Fable 5.1 and 45.5 percent for Opus 5.5. On AutomationBench, a test for business workflows, Fable 5.2 scores 51.4 percent, against 31.4 percent for Fable 5.1. Google's Gemini 4 Carbon does slightly better there, at 53.9 percent.

One independent check supports the picture: Fable 5.2 took first place on the Artificial Analysis Intelligence Index, with 65 points. As a research example, Anthropic says about 1,000 Fable 5.2 agents formalized the proof of Fermat's Last Theorem in Lean in about 30 hours.

Fable 5.2 is available in Claude paid plans, in Claude Code, on the Claude Platform, and through Amazon Web Services, Google Cloud and Microsoft Azure. GitHub Copilot and Cognition's coding tool Devin added it on the same day.

Fable 5.2 or Opus 5.5: which one to use

Anthropic's own advice is clear: start most coding and knowledge tasks with Claude Opus 5.5, and move to Fable for the most ambitious, open-ended technical work. In plain terms, that means Opus 5.5 for daily work and Fable 5.2 for the hard jobs. Large code refactors, long research tasks and multi-step agent work belong to Fable. Everything else stays with Opus.

For a small team, the message is: do not switch everything. Give Fable 5.2 the one or two jobs where depth really saves hours, and leave the rest where it is.

What Mythos 5.2 is, and why most businesses will never see it

Fable 5.2 and Mythos 5.2 are the same underlying model, according to Anthropic. Mythos has fewer cybersecurity and biology safeguards and is limited to vetted organisations, through Anthropic's Cyber Verification Program and Life Sciences Verification Program. Most businesses will never need it and should not ask for it. If a vendor promises you Mythos access, check carefully what they actually mean.

Three practical ways to use Fable 5.2 this week

First, long coding jobs. Give Fable a clearly described task for a whole codebase, with the goal, the limits and the checks. The one million token context window holds a lot of code at once, but a clear brief is still needed.

Second, long documents. Contracts, reports and studies can be summarised, compared and checked. Expect to verify every important statement yourself. The model saves reading time, it does not replace checking.

Third, agent workflows. Tasks with several steps, such as research, draft and check, gain the most from the higher speed. Thirty percent faster per step adds up across a whole workflow.

One worked example from the published prices: one million input tokens and 200,000 output tokens cost 16 US dollars. That is not a forecast for your bill, only a way to read the prices.

What to check before you switch

Check your tools. Claude Code, Copilot and Devin already support Fable 5.2. Check whether your own tools do too, before you make plans.

Test your most important prompts again. Anthropic says Fable 5.2 communicates more clearly than its predecessors. Do not take that as given. Try your three most important tasks with it first.

Set a budget. Cheaper per token does not mean cheaper overall, if the model invites heavier use. Set a weekly cap and look after two weeks at what was actually spent.

Keep a human in the loop. Anthropic writes openly that the new models voice suspicion of being evaluated more often than recent models. That makes safety testing harder to read. For your business it means: let AI prepare, but let people decide.

The takeaway

Fable 5.2 is a price cut and a speed gain on the model built for the hardest technical work. For most small businesses the right move is narrow: keep Opus 5.5 for daily work, give Fable 5.2 the one or two jobs where depth really pays, and measure the result in hours saved. For more on setting limits for agents, see our article on AI agents with a memory. For what new models mean for your website, see GPT-6 Astra and your website.