Claude Fable 5.2 is Anthropic's most capable model for general use, released on 9 November 2026. It costs $8 per million input tokens and $40 per million output tokens, 20% less than Fable 5.1, but still twice the price of Claude Opus 5.5. Anthropic itself recommends starting most coding and knowledge work with Opus 5.5 and moving up to Fable only for the hardest, most open-ended technical problems.
So the real question for most people is not "is it better?" (on Anthropic's own tests, clearly yes) but "is it worth twice the price for my work?". Below: the facts, what it did in testing, the benchmarks, how it differs from Mythos 5.2, and a simple way to decide.
Claude Fable 5.2 at a glance
All figures from Anthropic's announcement, model overview and pricing pages, checked on 10 November 2026.
| Claude Fable 5.2 | |
|---|---|
| Released | 9 November 2026 |
| API model ID | claude-fable-5-2 |
| Price | $8 input / $40 output per million tokens (Fable 5.1: $10 / $50) |
| Fast mode | up to 2.5x faster generation, $16 / $80 per million tokens |
| Prompt caching | cache reads $0.20; cache writes $10 (5 minutes) or $16 (1 hour) |
| Context window | 1 million tokens, up to 128,000 tokens per answer |
| Speed | up to 30% faster output than Fable 5.1, according to Anthropic |
| Where | Claude apps on paid plans, the Claude Platform API, AWS, Google Cloud, Microsoft Azure |
What it did before release
Anthropic published several examples from its own pre-release testing. They are the company's claims, not independent results, but they show what the model is aimed at: long, complicated jobs that run for hours or days.
- A famous proof, formalised in 30 hours. A team of 1,000 Fable 5.2 agents fully formalised the proof of Fermat's Last Theorem, turning it into a form a computer can check, in 30 hours with minimal human help. Anthropic's first formalisation, by an unreleased research version of Claude, took 11 days. Opus 5.5 in the same 1,000-agent setup completed 26% of the work in 18 hours.
- An operating system from an empty folder. Over eight days, coordinating more than 1,000 sub-agents, it wrote a prototype operating system that boots on a laptop: its own core, drivers for the disk, keyboard and network, file storage, windows, a login screen and an installer.
- Three working devices from one paragraph. With Diode, a circuit-board maker, it designed a home voice-assistant hub, a USB speakerphone and a handheld virtual pet: specifications, components, four-layer circuit boards, 3D-printed cases and software. Anthropic says all three worked without electrical modifications.
- Science. Anthropic says the model solved two open problems in the physics of black holes.
Benchmarks
Selected rows from the table in Anthropic's announcement. Higher is better. These are vendor-reported results and some use different effort settings, so treat them as a direction, not a final ranking.
| Test | Fable 5.2 | Fable 5.1 | Opus 5.5 | GPT-6 Astra | Gemini 4 Carbon |
| 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% |
| Remote Labor Index (real freelance-style tasks) | 41.3% | 17.9% | 21.3% | 20.8% | โ |
| RoboDojo Sim (robot control) | 64.9% | 27.6% | 39.4% | 29.8% | โ |
| Humanity's Last Exam, with tools | 74.2% | 65.6% | 67.7% | 57.2% | โ |
Three things stand out. The Remote Labor Index, which scores AI on paid remote work projects, nearly doubles against Opus 5.5 (41.3% vs 21.3%), and that is the closest of these tests to "can it do a job someone would pay for". Robot control more than doubles against GPT-6 Astra. And Fable 5.2 does not win everything: on AutomationBench, Google's Gemini 4 Carbon scores higher, though Carbon is not yet open to the public.
Fable 5.2 is not yet on the Arena leaderboard snapshot we use for our model catalogue, which dates from 27 October 2026, before its release.
Fable 5.2 vs Opus 5.5: is double the price worth it?
Opus 5.5 costs $4 / $20 per million tokens, half of Fable 5.2. The same job costs roughly twice as much on Fable, and it can cost more, because Fable's default reasoning effort is set to "high" while Opus 5.5 defaults to "medium", and reasoning is billed as output.
A simple example, our arithmetic at list prices, for one task that reads 50,000 tokens (a long document or a chunk of code) and writes 10,000:
| Model | Input cost | Output cost | Total per task | 1,000 tasks a month |
| Claude Opus 5.5 | $0.20 | $0.20 | $0.40 | $400 |
| Claude Fable 5.2 | $0.40 | $0.40 | $0.80 | $800 |
| Claude Fable 5.1 (previous) | $0.50 | $0.50 | $1.00 | $1,000 |
Pay for Fable when a task is long and open-ended and a failed attempt is expensive: a large code migration, a multi-day agent run, research where a missed step costs more than the tokens, hardware or science work. On those, a model that finishes the job in one run can be cheaper than a cheaper model that needs three.
Stay on Opus 5.5 (or Sonnet 5.5) when the work is routine: emails, documents, ordinary coding, data clean-up, customer-facing chat. Anthropic's own guidance is the same: start with Opus 5.5, go up to Fable or down to a smaller model depending on the job. Another pattern Anthropic suggests is using Fable as an adviser or coordinator over cheaper models, so the expensive model plans and checks while the cheap ones do the volume.
If you sell AI work to clients, this is the difference between a healthy margin and none. Price the job on the model it actually needs; the Claude money guide shows how to do that maths for agents and retainers.
Fable 5.2 vs Mythos 5.2
They are the same underlying model. Fable 5.2 carries strong safeguards for cybersecurity and biology so that Anthropic can offer it to everyone; Anthropic says these safeguards are as robust as Fable 5.1's but block fewer harmless requests. Mythos 5.2 has some of those safeguards lifted and is available only to vetted organisations: cybersecurity teams apply through Anthropic's Cyber Verification Program, life-sciences researchers through the Life Sciences Verification Program, and Claude Enterprise customers can use it inside Claude Security. Same price, $8 / $40.
For almost everyone, Mythos is not an option and not a need. The only reason to look at it is professional security or biomedical research.
How to get Claude Fable 5.2
| Plan | Fable access, per Anthropic's pricing page on 10 November 2026 |
| Free | not included |
| Pro | paid with usage credits; during launch week, included for up to 50% of weekly usage |
| Max and Ultra | included, up to half of your weekly limit |
| Team | standard seats included for up to 50% of weekly usage during launch week |
| API | claude-fable-5-2 on the Claude Platform, AWS, Google Cloud and Microsoft Azure |
During launch week Anthropic is also giving eligible paid subscribers one to three extra Fable resets, depending on how long they have subscribed. Fast mode is available in Claude Code and on the Claude Platform. If you are new to Claude's agent for building things, start with Claude Code for non-programmers.
FAQ
How much does Claude Fable 5.2 cost?
$8 per million input tokens and $40 per million output tokens through the API, 20% less than Fable 5.1. Fast mode costs $16 / $80. In the Claude apps it is included up to half the weekly limit on Max and Ultra, and paid with usage credits on Pro.
Is Claude Fable 5.2 better than Opus 5.5?
On Anthropic's published tests, yes, especially on long agent tasks, remote work projects and robot control. It also costs twice as much, and Anthropic recommends Opus 5.5 as the starting point for most coding and knowledge work.
What is the difference between Fable 5.2 and Fable 5.1?
Fable 5.2 is cheaper ($8/$40 vs $10/$50), up to 30% faster, and scores higher on every row of Anthropic's comparison table, by a wide margin on the Remote Labor Index (41.3% vs 17.9%) and robot control (64.9% vs 27.6%).
What is Claude Mythos 5.2, and can I use it?
The same model as Fable 5.2 with some cybersecurity and biology safeguards removed. Only vetted organisations get access, through Anthropic's verification programmes or Claude Security for Enterprise customers.
Is Claude Fable 5.2 available on the free plan?
No. It needs a paid plan or API access.
Sources
- Anthropic: Introducing Claude Fable 5.2 and Claude Mythos 5.2 (9 November 2026)
- Claude Platform: models overview and pricing (checked 10 November 2026)
- Claude plans and pricing (checked 10 November 2026)
