Anthropic released Claude Fable 5.2 on Monday, November 9, and the headline writes itself: the company’s most powerful generally available model is now faster, 20 percent cheaper, and ahead of every rival on the benchmarks Anthropic chose to publish. It is also, by the company’s own admission, not the model most people should use every day. Both of those things can be true at once.
Fable 5.2 arrives alongside Claude Mythos 5.2, which Anthropic describes as the same underlying model without some of Fable’s cybersecurity and biology safeguards, available only to vetted organizations. Fable is the one the rest of us get: on paid Claude plans, in Claude Code, on the Claude Platform as claude-fable-5-2, and through Amazon Web Services, Google Cloud, and Microsoft Azure.
What Fable 5.2 costs, and the new fast mode
The price is $8 per million input tokens and $40 per million output tokens, down from $10 and $50 for Fable 5.1. That is a straight 20 percent cut, and at current exchange rates it works out to roughly 700 rupees per million input tokens for Indian developers buying through the API. Anthropic also says the model writes up to 30 percent faster than Fable 5.1, and it adds an optional fast mode that runs up to 2.5 times faster still, priced at $16 and $80 per million tokens.
That gives developers three dials to turn instead of one: reasoning effort, speed, and cost per task. Anthropic’s system card makes the tradeoff explicit. At High effort, Fable 5.2 matched Claude Opus 5.5‘s best score on the FrontierCode coding test, 45.3 percent against 45.5 percent, for roughly half the cost per task. At maximum effort it scores higher still, but the cost per task jumps to $44.70, more than triple the High effort figure. The cheap setting is not a cut-down version of the model. It is the same model, spending less time thinking, and for a surprising number of jobs that is enough.
The benchmark table: big jumps in coding, one clear loss
Anthropic’s own scores put Fable 5.2 ahead of Fable 5.1 on every test it published, and ahead of OpenAI’s GPT-6 Astra and Google’s Gemini 4 Carbon on most of them. These are vendor-reported numbers, run by Anthropic or by evaluators it selected, and none of them has been independently replicated yet. Read them as a claim, not a verdict. Here are the rows that matter most:
| Benchmark | Fable 5.2 | Fable 5.1 | Opus 5.5 | GPT-6 Astra | Gemini 4 Carbon |
|---|---|---|---|---|---|
| FrontierCode 2.0 (agentic coding) | 49.1% | 32.4% | 45.5% | 42.8% | 41.1% |
| FrontierSWE v2 (agentic coding) | 69.9% | 56.3% | 62.3% | 65.5% | 61.2% |
| AutomationBench (business workflows) | 51.4% | 31.4% | 42.5% | 41.4% | 53.9% |
| OSWorld 2.1 (computer use) | 89.3% | 83.5% | 87.2% | 79.5% | 81.5% |
| Humanity’s Last Exam (with tools) | 74.2% | 65.6% | 67.7% | 57.2% | Not given |
| Remote Labor Index | 41.3% | 17.9% | 21.3% | 20.8% | Not given |
The coding jump over Fable 5.1 is the story: FrontierCode 2.0 rises 16.7 points, FrontierSWE v2 rises 13.6 points, and the Remote Labor Index more than doubles, from 17.9 to 41.3 percent. But Anthropic itself flags the exceptions. Google’s Gemini 4 Carbon wins AutomationBench, the business workflow test, 53.9 percent to 51.4. Opus 5.5 still leads on the AA-Briefcase knowledge-work Elo rating, 1800 to 1785. And on Terminal-Bench 4.0, Fable 5.2 scores 60 percent, level with Opus 5.5 and behind Kimi K3.1 Pro, Gemini 4 Carbon, and Claude Sonnet 5.5. The honest summary is Anthropic’s own line: Fable 5.2 is “a step above Opus 5.5 in almost all the capabilities we evaluated.” Almost, not all.
The independent numbers look good, with a cost warning
Two outside labs have published results already, and both put Fable 5.2 in first place. Vals AI ranks it first of 36 models on its Vals Index, at 71.31 percent, ahead of Claude Sonnet 5.5 at 70.85 percent and Gemini 4 Carbon at 70.45 percent, with a margin of error of about a point. Artificial Analysis puts Fable 5.2 at maximum effort first on its Intelligence Index with a score of 65, ahead of Gemini 4 Carbon at 60, Opus 5.5 at 59, and GPT-6 Astra at 57.
Artificial Analysis also publishes the catch. At maximum effort, running the index cost an average of $26.54 per task, about 2.6 times as much as Opus 5.5, and at 44.1 minutes per task it rates Fable 5.2 among the slowest models in its price range. At the High and extra-high effort settings it still scored 62, which beats every other model the firm has tested, while using fewer output tokens per task than Opus 5.5 at maximum effort. The lesson is the same one Anthropic’s own cost table teaches: the top score is real, and it is expensive. Most of the value sits one or two effort settings below the top.
What early testers built, and what to make of it
Anthropic says it built a working operating system with the model in eight days, and early testers have described projects including a website of more than 450 generated 3D worlds, a browser spreadsheet that took 50 hours and roughly $11,700 of tokens, and a custom video editor. Treat these as demonstrations, not evidence. Several testers used the model free of charge, and most of the results have not been checked independently. What they do show is the shape of the work Fable is built for: long, multi-hour technical tasks where the model keeps going after a human would have stopped to sleep.
Free Fable resets for launch week
For subscribers, Anthropic is running a launch-week offer. Pro and standard Team seats get access to all Fable models until November 16, for up to 50 percent of their weekly limits, plus one Fable limit reset. Max, Ultra, and premium seats get one to three resets depending on how long they have paid, usable until March 8, 2027. If you have been rationing Fable usage on a Pro plan, this is the week to try the jobs you have been putting off.
Meta’s Muse Ember joins the bargain fight
Fable 5.2 was not the only model launch on Monday. Meta also shipped Muse Ember, a small and very cheap model that matches GPT-6 Luna’s API pricing at $0.10 per million input tokens and $0.50 per million output tokens, while scoring 36 on the Artificial Analysis Intelligence Index, matching Claude Haiku 5.5 at maximum effort for about a third of the cost per task. The frontier and the bargain bin moved on the same day, and the bargain model may matter more to most builders.
Should you switch to Fable 5.2?
If your work is long, agentic coding or multi-hour technical tasks, yes, and start at High effort rather than maximum: you get nearly the top score for roughly half the cost per task. If your work is everyday writing, analysis, and coding help, Anthropic still recommends Opus 5.5, and the benchmarks back that up: Opus wins the knowledge-work Elo rating, costs far less per task at its own maximum effort, and finishes faster. Fable 5.2 is the strongest general model Anthropic has shipped, and the honest way to use it is to spend its power where the job is hard enough to need it.
