Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 on November 9, cutting Claude Fable 5.2 pricing 20% to $8 per million input tokens and $40 per million output tokens. That undercuts OpenAI’s GPT-6 Astra at $10 and $50. Anthropic says the model generates output up to 30% faster than Fable 5.1 and beats Opus 5.5 on almost every test it ran.
Key takeaways
- Fable 5.2 lists at $8 input and $40 output per million tokens, 20% below Fable 5.1.
- Cache hits cost $0.20 per million tokens, against $1.00 on GPT-6 Astra.
- Anthropic says Fable 5.2 scored 49.1% on FrontierCode 2.0, versus 45.5% for Opus 5.5.
What is Claude Fable 5.2?
Claude Fable 5.2 is Anthropic’s most capable public model, built for long, open-ended technical work such as code migrations, hardware design and research. It shares one underlying model with Claude Mythos 5.2. Fable adds cybersecurity and biology safeguards so anyone can use it, while Mythos goes only to vetted organizations.
Anthropic announced both models on November 9. Developers call Fable 5.2 as claude-fable-5-2, and it launched the same day on the Claude API, Amazon Web Services, Google Cloud and Microsoft Azure, according to Anthropic’s announcement.
It is Anthropic’s first frontier release since CEO Dario Amodei called in September for the industry to slow AI progress, CNBC reported. Dianne Penn, Anthropic’s head of product management for research, told CNBC: “We called for pacing, not pausing.”
The demo built for headlines: an Anthropic engineer gave Fable 5.2 an empty folder and asked for an operating system. Over eight days, the model coordinated more than 1,000 subagents and produced a system that boots on a laptop, with drivers, file storage, windows and a login screen.
Read the fine print, though. Anthropic itself tells most users to start with Claude Opus 5.5 and move up to Fable only for the hardest work. The flagship is a specialist tool, not a default.
Claude Fable 5.2 pricing: what does it cost now?
Claude Fable 5.2 costs $8 per million input tokens and $40 per million output tokens on the Claude API, down from $10 and $50 for Fable 5.1. Cache hits drop to $0.20 per million, batch jobs cost half, and an optional fast mode runs $16 and $80 per million tokens.
| Model | Input / 1M | Output / 1M | Cache hit / 1M | Fast mode (in / out) |
|---|---|---|---|---|
| Claude Fable 5.2 | $8 | $40 | $0.20 | $16 / $80 |
| Claude Fable 5.1 | $10 | $50 | $0.25 | Not listed |
| Claude Opus 5.5 | $4 | $20 | $0.20 | $8 / $40 |
| GPT-6 Astra | $10 | $50 | $1.00 | $20 / $100 |
On list price, Fable 5.2 is now 20% cheaper than GPT-6 Astra, which still shows $10 and $50 on OpenAI’s pricing page. The bigger gap is cached input: $0.20 against Astra’s $1.00.
That 80% difference matters most for agents that reread the same system prompt, tool schemas and codebase context hundreds of times in one job.
What does a real workload cost?
Take 10 million input tokens and 2 million output tokens a day, with no caching. By our arithmetic, Fable 5.2 costs $160 a day, or $4,800 over 30 days.
Fable 5.1 and GPT-6 Astra both cost $200 a day on the same load, or $6,000 a month. Opus 5.5 costs $80 a day.
Batch and fast mode
Anthropic’s Batch API halves Fable 5.2 to $4 and $20 for jobs that can wait, according to Anthropic’s pricing docs. Fast mode, which Anthropic says speeds token generation by up to 2.5 times, doubles the standard rate and cannot be combined with batch.
One caution on per-token math. The same docs say Anthropic’s newer tokenizer produces about 30% more tokens for the same text than its older one. Vendors count tokens differently, so a lower list price does not guarantee a lower bill.
Is Claude Fable 5.2 better than GPT-6 Astra?
On Anthropic’s own tests, mostly yes. Fable 5.2 beats GPT-6 Astra on agentic coding, computer use, robotics and Humanity’s Last Exam. But every number below is vendor-reported, Google’s Gemini 4 Carbon wins one business-workflow test, and our searches found no independent leaderboard score for Fable 5.2 yet.
| 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 | 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 | 41.3% | 17.9% | 21.3% | 20.8% | n/a |
| Humanity’s Last Exam (with tools) | 74.2% | 65.6% | 67.7% | 57.2% | n/a |
| RoboDojo Sim (robotics) | 64.9% | 27.6% | 39.4% | 29.8% | n/a |
Google’s model leads AutomationBench at 53.9% against 51.4% for Fable 5.2. For more on that rival, see our Gemini 4 Carbon launch coverage.
Why cost per task matters more than list price
Anthropic’s sharper claim is about efficiency. At High effort, Fable 5.2 scored 45.3% on FrontierCode 2.0, level with Opus 5.5’s best of 45.5%, for roughly half the cost per task, the company said.
On Humanity’s Last Exam, Fable 5.2 at Low effort matched Opus 5.5’s top score for less than one-tenth of the cost per task. If those ratios hold outside Anthropic’s own harness, the pricier model can be the cheaper one per finished job.
They may not. Opus 5.5 still edges Fable 5.2 on the AA-Briefcase knowledge-work Elo, 1,800 to 1,785, in Anthropic’s own table.
What is Claude Mythos 5.2, and who can use it?
Claude Mythos 5.2 is the same model with some cybersecurity and biology safeguards removed. It costs the same $8 and $40, but only organizations admitted to Anthropic’s Cyber or Life Sciences Verification Programs can use it. Claude Enterprise customers can also reach it through Claude Security, Anthropic’s code-scanning product.
On Anthropic’s internal CVE Rediscovery test, run in October against 1,829 published vulnerabilities, Mythos 5.2 found 10.2% more than Opus 5.5 and 14.4% more than Mythos 5.1.
Fable 5.2 blocks most of that work. Blocked cyber requests fall back to Opus 4.8 and blocked biology requests to Opus 5. Anthropic says early testing showed about 55% fewer blocks than Fable 5.1.
Who wins and who loses from the Fable 5.2 price cut?
Developers and enterprise buyers win a cheaper top-tier model, and the three big clouds win another product to resell. OpenAI loses price parity at the top of the market. Anthropic trades 20% of revenue per token for volume, just as public investors start to price the company.
- Winner: teams running coding agents and long automations. List prices fall 20%, and cache hits fall from $0.25 to $0.20 per million tokens versus Fable 5.1.
- Winner: Amazon, Google and Microsoft. Fable 5.2 shipped on AWS, Google Cloud and Azure on day one, so cheaper tokens mean more traffic through their clouds.
- Loser: OpenAI. GPT-6 Astra now lists 25% above Fable 5.2 on input and output, and five times higher on cache hits. OpenAI’s recent cuts, such as GPT-6.1 Terra at $1 and $5, sit in the middle of its lineup, not the top.
- Mixed: Anthropic’s own Opus 5.5. At $4 and $20 it costs half as much as Fable, but if Fable matches Opus at half the cost per task, some Opus spend moves up. The revenue stays in-house either way.
The investor read is less tidy. A 20% cut means each token earns 20% less, so the move pays only if usage grows faster than prices fall.
Anthropic made the cut as CNBC reported its IPO could come as soon as next week. Our breakdown of the Anthropic IPO price range covers the deal terms.
The price war is running at the bottom of the market too. Meta launched Muse Ember on November 9 at $0.10 per million input tokens and $0.50 output, the same list price as Claude Haiku 5.5 and GPT-6 Luna, Adgully reported.
What risks does Anthropic disclose?
Anthropic says high-stakes misalignment risk stays low, but it flags a growing problem: both models voice suspicion that they are being evaluated more often than any recent Claude model. That weakens lab tests as a guide to real-world behavior, and Anthropic expects the problem to grow with future models.
In Anthropic’s words, that suspicion “challenges our ability to assess how they will act in real-world settings.”
Anthropic also says it deliberately held AI-research skills, such as accelerator kernel optimization, near Opus 5.5 levels. That matters after the company’s filing warned it may soon hit its own threshold for AI-accelerated progress, as we covered in Anthropic’s recursive self-improvement warning.
For enterprise buyers, data handling is the practical catch. Most customers get 30-day retention for safety monitoring. Zero data retention requires Enterprise Frontier Safeguards, which is still in private beta.
What to watch next
Three things will decide whether this cut reshapes the market: how Anthropic’s IPO prices, whether OpenAI answers on GPT-6 Astra pricing, and what independent benchmarks show. Below are the dated events and concrete signals to track, with dates only where a company or a credible report has set them.
- This week: Anthropic’s Fable 5.2 launch promotion gives eligible paid subscribers extra Fable usage. Watch whether limits tighten once it ends.
- As soon as next week: CNBC reported the IPO could happen as early as next week. The final price will show how much credit investors give the model roadmap.
- November 30: OpenAI’s promotional pricing on GPT-5.6 Sol expires, per its pricing page. It is the next scheduled change on OpenAI’s rate card.
- Signal: independent scores. Our searches found no Artificial Analysis Intelligence Index result for Fable 5.2 yet. The first third-party cost-per-task figures will test Anthropic’s half-the-cost claim.
FAQ
How much does Claude Fable 5.2 cost?
$8 per million input tokens and $40 per million output tokens on the Claude API. Cache hits cost $0.20 per million, batch jobs cost $4 and $20, and fast mode costs $16 and $80.
Is Claude Fable 5.2 cheaper than GPT-6 Astra?
On list price, yes. GPT-6 Astra costs $10 input, $50 output and $1.00 for cached input per million tokens. Real bills depend on how many tokens each model uses per task.
What is the difference between Claude Fable 5.2 and Claude Mythos 5.2?
They are the same underlying model at the same price. Fable 5.2 has cybersecurity and biology safeguards and is open to everyone; Mythos 5.2 drops some safeguards and is limited to vetted organizations.
When was Claude Fable 5.2 released?
November 9, 2026. It launched on the Claude API, AWS, Google Cloud and Microsoft Azure under the model ID claude-fable-5-2.
Should I use Claude Fable 5.2 or Claude Opus 5.5?
Anthropic recommends starting with Opus 5.5, at $4 and $20 per million tokens, for most coding and knowledge work, and moving up to Fable 5.2 for the most open-ended technical tasks.
The bottom line
Fable 5.2 is the rare frontier launch where price moves in the buyer’s favor: more capability, 20% less per token and cache reads 80% cheaper than GPT-6 Astra’s. For teams already on Fable 5.1, switching is an easy call.
For everyone else, Anthropic’s own advice holds: start on Opus 5.5 at half the price and escalate only when a task demands it. For investors, the cut is a volume bet timed to an IPO. The number to watch next is OpenAI’s price on its next Astra.
Free guide · 12 pages
AI Tools Every Investor Should Use in 2026
10 tools for research, stock signals, charts and crypto, what each one costs, three ready-made stacks from $0 a month and 7 copy-paste prompts for filings and earnings calls. Enter your email and we will send it to you.
Plus one email a week: AI Money This Week, every Sunday. Unsubscribe anytime.
Sources
- Anthropic: Introducing Claude Fable 5.2 and Claude Mythos 5.2 (November 9, 2026)
- Anthropic: Claude Platform pricing (checked November 10, 2026)
- OpenAI: API pricing (checked November 10, 2026)
- CNBC: Anthropic releases Claude Fable 5.2 and Claude Mythos 5.2 (November 9, 2026)
- Adgully: Meta debuts low-cost AI model Muse Ember to power agentic subtasks (November 10, 2026)
Wealth Engine researches and drafts with AI tools and checks every figure against the sources above. How we report.
Leave a Reply