Short answer: Claude Fable 5.2 is Anthropic's most capable model available to everyone, released on 9 November 2026. It writes output up to 30% faster than Fable 5.1, costs 20% less ($8 per million input tokens and $40 per million output tokens, down from $10 and $50), and scores higher than Claude Opus 5.5 on most of Anthropic's published benchmarks. It's built for long, difficult technical work. For everyday business tasks, Anthropic itself recommends starting with Opus 5.5 and moving up to Fable only when the job needs it.
We build AI automation for clients and use Claude models every day, so this guide is written for business owners and teams deciding whether Fable 5.2 is worth paying for. All facts and figures come from Anthropic's official announcement and its developer documentation.
What is Claude Fable 5.2?
Claude Fable 5.2 is the newest model in Anthropic's top tier, above the Opus, Sonnet and Haiku lines. It's designed for what Anthropic calls the most daunting technical challenges: large software projects, multi-day agent tasks, research and engineering problems.
It launched alongside Claude Mythos 5.2. According to Anthropic, Fable and Mythos are the same underlying model. The difference is access:
- Fable 5.2 has strong safeguards around cybersecurity and biology, which lets Anthropic offer it to all users.
- Mythos 5.2 is for verified organisations doing beneficial cybersecurity or biomedical work, through Anthropic's Cyber Verification and Life Sciences programmes.
For almost every business, Fable 5.2 is the one you'll use.
What's new in Fable 5.2 compared with Fable 5.1
| Fable 5.1 | Fable 5.2 | |
|---|---|---|
| Input price (per 1M tokens) | $10 | $8 |
| Output price (per 1M tokens) | $50 | $40 |
| Output speed | Baseline | Up to 30% faster (plus an optional fast mode, up to 2.5x faster at $16 / $80) |
| Context window | 1M tokens | 1M tokens |
| FrontierCode 2.0 (agentic coding) | 32.4% | 49.1% |
| AutomationBench (business workflows) | 31.4% | 51.4% |
Beyond the numbers, Anthropic says testers found Fable 5.2 communicates more clearly than earlier Fable models, and that its safety filters block fewer legitimate requests than Fable 5.1's while staying as strict on real misuse.
How much does Claude Fable 5.2 cost?
On the Claude API, Fable 5.2 costs $8 per million input tokens and $40 per million output tokens. Cached input reads cost $0.20 per million. The optional fast mode costs $16 and $80.
Tokens are pieces of words; a rough rule of thumb is that 1,000 tokens is about 750 English words. To give a feel for the price, here's what one request with 50,000 tokens in (for example, a long document plus instructions) and 10,000 tokens out (a detailed report) would cost on each current model:
| Model | Price per 1M tokens (in / out) | Cost of that one request |
|---|---|---|
| Claude Fable 5.2 | $8 / $40 | $0.80 |
| Claude Opus 5.5 | $4 / $20 | $0.40 |
| Claude Sonnet 5.5 | $2 / $10 | $0.20 |
| Claude Haiku 5.5 | $0.10 / $0.50 | $0.01 |
So per request, Fable 5.2 costs twice as much as Opus 5.5. The per-task picture can be different, though. Anthropic reports that on its FrontierCode 2.0 coding test, Fable 5.2 at High effort scored about the same as Opus 5.5 at its maximum setting for roughly half the cost per task, because it needed fewer tokens to get there. On hard problems, the more expensive model can be the cheaper way to finish the job.
If you use Claude through a subscription rather than the API, Anthropic says eligible paid plans include Fable, with extra Fable usage during launch week.
How Fable 5.2 compares with Opus 5.5 and other models
Here are some of the results from Anthropic's published benchmark table (scores are Anthropic's own measurements):
| Test | Fable 5.2 | Opus 5.5 | GPT-6 Astra | Gemini 4 Carbon |
|---|---|---|---|---|
| FrontierCode 2.0 (agentic coding) | 49.1% | 45.5% | 42.8% | 41.1% |
| AutomationBench (business workflows) | 51.4% | 42.5% | 41.4% | 53.9% |
| OSWorld 2.1 (computer use) | 89.3% | 87.2% | 79.5% | 81.5% |
| Humanity's Last Exam (with tools) | 74.2% | 67.7% | 57.2% | n/a |
| Remote Labor Index (knowledge work) | 41.3% | 21.3% | 20.8% | n/a |
Two things stand out for business owners. First, Fable 5.2 nearly doubles Opus 5.5 on the Remote Labor Index, a test built from real paid work projects. Second, it doesn't win everywhere: on AutomationBench, Google's Gemini 4 Carbon scored slightly higher. No single model is best at everything, which is why we test models on each client's real tasks before choosing (see our ChatGPT vs Claude vs Gemini comparison).
What businesses can use Claude Fable 5.2 for
Anthropic's examples are dramatic (an agent team migrated a production service from Python to Rust and made it start about 26 times faster; another run formalised the proof of Fermat's Last Theorem in 30 hours). Translated into ordinary business work, these are the jobs where the extra cost makes sense:
- Large software projects and migrations. Moving an old system to a new stack, upgrading a codebase, or building a feature that touches many parts of an app. This is where Fable's coding lead matters most.
- Complex reporting and SQL. Anthropic shares an Oracle test where Fable 5.2 produced correct SQL for a multi-table financial calculation while two other models returned numbers 5 to 8 times too high. When a wrong figure is costly, accuracy is worth paying for.
- Long, unattended agent tasks. Audits, research reports and multi-step workflows that run for hours. Testers describe it working until the job is done rather than stopping at a plausible checkpoint.
- Designing an automation system. Planning how an AI system should work (which steps, which tools, which checks) before cheaper models run it day to day.
- As an "advisor" for cheaper models. Anthropic notes that Fable 5.2 can act as an advisor that Haiku, Sonnet or Opus consult on hard problems, or as an orchestrator that hands simpler tasks to smaller models. That gives you top-tier judgment at a fraction of the cost.
When not to use Fable 5.2
Most business AI work doesn't need the top model:
- Website chatbots and WhatsApp replies: Sonnet 5.5 or Haiku 5.5 are faster and far cheaper.
- High-volume tasks like sorting enquiries, tagging leads or summarising emails: Haiku 5.5 costs about 1/80th of Fable 5.2 per token.
- Everyday writing, analysis and coding: Anthropic recommends Opus 5.5 as the daily default.
- Anything needing zero data retention: Fable 5.2 uses 30-day retention for safety monitoring unless your organisation is eligible for Anthropic's Enterprise Frontier Safeguards. Check this before sending sensitive data.
A simple rule we use with clients: start with the cheapest model that does the job well, and only move up when you can see the quality gap on your own examples.
How to access Claude Fable 5.2
- Claude apps: on eligible paid plans, choose Fable 5.2 in the model menu.
- Claude API: use the model ID
claude-fable-5-2. - Cloud platforms: Anthropic says it's available on all platforms, including Amazon Web Services, Google Cloud and Microsoft Azure.
- Claude Code: the coding tool can use Fable for big projects. Our Claude Code guide explains how it works.
For developers: thinking is always on in Fable 5.2, and you control how hard it works with the effort setting (low, medium, high, xhigh or max). Anthropic's docs suggest starting at high and only going above it when you can measure a gain. Medium is the default in Claude's own apps.
Should your business use Claude Fable 5.2?
If you're building or migrating software, running long AI agent workflows, or need numbers you can trust in complex reports, Fable 5.2 is now cheaper and faster than the model it replaces and is worth testing. For chatbots, everyday content and high-volume automation, Opus 5.5, Sonnet 5.5 or Haiku 5.5 will give you better value.
Not sure which model fits your workflow? Our AI automation team tests the options on your real tasks and builds the system around the best value, often combining a top model for planning with cheaper models for the daily work. If you want to connect Claude to your own tools, our guide to MCP is a good next read.
