Claude Fable 5.2: price, access and what's new

Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 on Monday, November 9, 2026. They are new versions of its most powerful AI models, built for the hardest technical work: big coding projects, scientific research and long tasks that run for hours or days.

The headline for most people is simple. Fable 5.2 is available today to everyone who can already use Fable, it is up to 30% faster than Fable 5.1, and it costs 20% less to use through the API. Mythos 5.2, the same model with fewer safety blocks for security and biology work, stays limited to vetted organizations. Here is what changed, what it costs, who can use it and how it compares with OpenAI's GPT-6 Astra and Google's Gemini 4 Carbon.

What Anthropic announced

In its announcement, Anthropic says Fable 5.2 and Mythos 5.2 "show improvements in areas like coding, reasoning, and vision." The two follow Fable 5.1 and Mythos 5.1, which arrived in September.

Fable and Mythos are the same underlying model. The difference is safeguards. Fable has strong cybersecurity and biology filters that block high-risk requests, such as writing exploits or penetration testing, which lets Anthropic offer it to all users. Mythos has those filters relaxed, so it is only for organizations approved through Anthropic's Cyber Verification Program or Life Sciences Verification Program.

Anthropic also says testers found the new models "communicate more clearly" than earlier Fable and Mythos versions, and that they produce answers up to 30% faster than the 5.1 models.

Claude Fable 5.2 price

Through the API, Fable 5.2 and Mythos 5.2 cost $8 per million input tokens and $40 per million output tokens, according to Anthropic's announcement and model documentation. That is down from $10 and $50 for the 5.1 models. Reading from the prompt cache drops to $0.20 per million tokens, from $0.25.

There is also an optional fast mode that Anthropic says speeds up output by up to 2.5 times. It costs double: $16 per million input tokens and $80 per million output tokens. Batch requests get a 50% discount. Both models have a 1 million token context window and can write up to 128,000 tokens in one response.

For comparison, OpenAI's own pricing page lists GPT-6 Astra at $10 per million input tokens, $50 per million output tokens and $1 for cached input. Fortune points out that this makes Fable 5.2 about 20% cheaper on normal input and output, and 80% cheaper on cached input, which matters when you send the same long documents again and again.

Prices checked against Anthropic's and OpenAI's official pages on November 9, 2026 at 21:25 UTC (November 10, 02:55 IST).

Who can use it, and the launch offer

Fable 5.2 is live now in the Claude apps, in Claude Code and on the Claude Platform (model name claude-fable-5-2). It is also on Amazon Bedrock, Google Cloud and Microsoft Foundry on Azure. Amazon confirmed availability on Bedrock on launch day.

For subscribers, Anthropic is running a launch promotion. According to its help center:

Pro plans and standard Team seats: all Fable models are included in the plan from November 9 to November 16, 2026 at 11:59 pm PT, for up to 50% of the weekly usage limit, plus one Fable limit reset. After that week, Fable goes back to running on usage credits for these plans.

Max and Ultra plans, premium Team seats and premium seats on legacy seat-based Enterprise plans: one to three Fable limit resets, depending on how long you have paid for Claude. Under six months gets one, six to 12 months two, and more than 12 months three. These resets expire on March 8, 2027.

Free accounts are not included. A reset gives back the usage you have already spent on Fable models in your weekly limit, and refills the five-hour limit if your account has one. It does not raise your limits. You use a reset from Settings, then Usage. Your default model also does not change, so you have to pick Fable from the model picker.

Mythos 5.2 works differently. Security teams can apply through the Cyber Verification Program and life sciences researchers through the Life Sciences Verification Program. Claude Enterprise customers can also reach it indirectly through Claude Security, Anthropic's code-scanning product.

Benchmarks: how it compares

Anthropic published a table comparing Fable 5.2 with Fable 5.1, its own Opus 5.5, GPT-6 Astra and Gemini 4 Carbon. These are company-run results, so treat them as claims until independent testers weigh in. Highlights:

Coding: on FrontierSWE v2, Fable 5.2 scores 69.9%, against 65.5% for GPT-6 Astra, 62.3% for Opus 5.5 and 61.2% for Gemini 4 Carbon. On FrontierCode 2.0, which checks whether code changes would be accepted by real open-source projects, it reaches 49.1% at its highest effort setting, versus 42.8% for GPT-6 Astra.

Computer use: 89.3% on OSWorld 2.1, ahead of Opus 5.5 (87.2%), Gemini 4 Carbon (81.5%) and GPT-6 Astra (79.5%).

Reasoning: 74.2% on Humanity's Last Exam with tools, against 57.2% for GPT-6 Astra.

Robotics: 64.9% on the RoboDojo simulated robot test, more than double GPT-6 Astra's 29.8% in Anthropic's own runs.

Not every result is a win. On AutomationBench, a business-workflow test, Gemini 4 Carbon scores 53.9% against Fable 5.2's 51.4%. And Anthropic still recommends Opus 5.5 as the everyday model for most coding and knowledge work, with Fable kept for the most ambitious jobs.

What Fable 5.2 did in testing

Anthropic listed several examples from before launch. It says Fable 5.2 solved two open problems in theoretical physics about black holes, and that a team of about 1,000 Fable 5.2 agents translated the proof of Fermat's Last Theorem into a form a computer can check, in 30 hours. An earlier research version of Claude took 11 days.

In another test, an Anthropic engineer asked Fable 5.2 to build a prototype operating system from an empty folder. Anthropic says the model coordinated more than 1,000 sub-agents over eight days and produced a system that boots on a laptop, with drivers, file storage, windows and an installer.

The most tangible example is hardware. From a one-paragraph prompt, Anthropic says Fable 5.2 designed three working gadgets with their own circuit boards, 3D-printed cases and software: a home voice-assistant hub, a USB speakerphone and a handheld virtual pet. Circuit board maker Diode built the boards, and Anthropic says all three worked without any electrical fixes.

Anthropic also says over 90% of the code changes merged into its own main codebase are now written by Claude, with human guidance, and that a team of agents running an early Fable 5.2 moved a key service from Python to Rust, making it start about 26 times faster.

Safety: fewer blocks, an odd new problem

Anthropic says Fable 5.2's security and biology filters are as strong as before but block legitimate requests less often: about 55% fewer blocks than Fable 5.1 in early testing. When a request is blocked, the answer comes from an older Opus model instead.

The company also flagged a testing problem. In its words, Mythos 5.2 and Fable 5.2 "each voice suspicion that they are being evaluated more often than any of our recent models," which makes it harder to know how they will behave in real use. TechCrunch notes this is Anthropic's first model release since a report from the independent evaluation group METR suggested Anthropic's internal models were speeding up its own AI progress by roughly two times. Anthropic says outside testers, including the UK AI Security Institute and METR, checked the model before release, and that it judges the risk of seriously unsafe behavior to be low.

On data, both models require 30-day data retention by default. Anthropic is offering a private beta called Enterprise Frontier Safeguards so that some business customers can run the safety monitoring on their own systems and keep zero data retention.

Should you switch?

If you already pay for Max or Ultra, or you are on Pro this week, it costs nothing extra to try Fable 5.2 from the model picker, and the launch resets give you room to test it on a real project. Keep an eye on Settings, then Usage, because Anthropic warns Fable models use up limits faster than its other models.

For developers, the price cut makes Fable 5.2 cheaper than GPT-6 Astra per token, but it is still far more expensive than everyday models. Anthropic's own advice is to start with Opus 5.5 and step up to Fable only for the hardest problems. The real test will be independent benchmarks and day-to-day use over the next few weeks.