Anthropic Launches Claude Fable 5.2 and Mythos 5.2, Cuts Price 20% as IPO Nears

Anthropic has released Claude Fable 5.2 and Claude Mythos 5.2, its newest flagship AI models, on 9 November 2026 — and cut the price by 20 percent at the same time. Both models cost $8 per million input tokens and $40 per million output tokens on the Claude API, down from $10 and $50 for the 5.1 versions. They arrive with a 1-million-token context window, up to 128,000 tokens of output, and benchmark scores that put Fable 5.2 at the top of at least one independent ranking — just as Anthropic prepares for a stock market listing that could value the company at more than a trillion dollars.

The two names describe one model with two different safety settings. Fable 5.2 is the version for general users and developers. Mythos 5.2 is the same underlying model with fewer restrictions, available only to organisations that pass Anthropic’s Cyber Verification Program or its Life Sciences Verification Program. The split matters because the most powerful version of the model is deliberately kept away from general use — and because the price cut applies to both.

One model, two safeguard levels

Anthropic is explicit that Fable 5.2 and Mythos 5.2 are not two models of different sizes or training. They share the same underlying model, the same context length, the same output limits and the same pricing. What differs is access and safety.

Fable 5.2 is available to all API customers and to eligible paid Claude subscribers. It blocks most high-risk cybersecurity work — penetration testing, exploit attempts, binary vulnerability analysis — and most research-biology requests. Mythos 5.2 drops those blocks for vetted organisations working in security and the life sciences. That is why the two versions exist: the same capability, with the guardrails set differently for different users.

There are practical consequences. Fable 5.2 requires a default 30-day data retention period, and it does not support zero data retention unless Anthropic specifically authorises it. Mythos 5.2 is not part of the Pro, Max or Ultra plans at all — verified organisations reach it through Anthropic’s verification programs, and Enterprise customers can reach its scans through Claude Security. And despite the new release, Anthropic still recommends Opus 5.5 as the everyday default model, not Fable or Mythos.

The price cut — and the faster, dearer mode

The headline change is price. Standard API pricing is now $8 per million input tokens and $40 per million output tokens, a 20 percent reduction on every price component compared with Fable 5.1. Anthropic also says standard-mode output is up to 30 percent faster than the 5.1 versions.

There is also a new optional fast mode, which generates tokens up to 2.5 times faster — at double the price: $16 per million input tokens and $80 per million output tokens. Cache reads cost $0.20 per million tokens, and batch processing reduces the base input and output prices by 50 percent. The full 1-million-token context window is billed at the standard per-token rate, so a very long request carries no surcharge for the context itself.

For developers, the maths is straightforward: the same work that cost $10 and $50 per million tokens last week now costs $8 and $40, and the faster mode is there when speed matters more than cost. For Anthropic, the cut is a competitive move — the company is pricing its flagship below its predecessor at the same time as its rivals are cutting their own prices.

What the model can do

The specifications are unchanged from the 5.1 generation in size but improved in performance. The context window is 1 million tokens. Maximum output is 128,000 tokens. The knowledge cutoff is June 2026. Adaptive thinking is always on, and the API default effort is high. The model IDs are claude-fable-5-2 and claude-mythos-5-2 (on Amazon Bedrock, anthropic.claude-fable-5-2). It is available from launch on the Claude API, Amazon Bedrock, Claude Models on Google Cloud, and Microsoft Foundry. Anthropic commits to keeping the model available until at least 9 November 2027.

On Anthropic’s own benchmark, Fable 5.2 scored 49.1 percent on FrontierCode 2.0, a coding benchmark, against 32.4 percent for Fable 5.1 and 45.5 percent for Opus 5.5. In a pre-release experiment, a team of about 1,000 Fable 5.2 agents formalised Fermat’s Last Theorem in the Lean proof assistant in about 30 hours. That was Anthropic’s own internal experiment, designed by the company, and it should not be read as the model independently inventing new mathematics — but it is a measure of how far multi-agent work with the model has come.

The independent check is more interesting. Artificial Analysis, in its VIALS-AA evaluation, puts Fable 5.2 first of 21 models at 67.1 percent, ahead of Gemini 4 Carbon at 64.0 percent and GPT-6 Astra at 62.7 percent. But the same evaluation shows the cost of that lead: Fable 5.2 used 37.4 tool calls per task, 4.7 times as many as GPT-6 Astra — and every one of those calls lands on the bill. The model also did not lead in legal, tax, or certain multimodal programming tasks. Top reasoning performance, in other words, still costs significant time and money.

Who gets it, and the launch-week offer

Fable 5.2 is available on all paid Claude plans — Pro, Max, Ultra, Team and Enterprise — but not on the Free plan. Pro costs $20 a month and Max costs $100 a month at standard Claude prices.

For the launch week, Pro subscribers and standard Team seats get Fable models included for up to 50 percent of their weekly usage limits, from 9 November through 16 November 2026 at 11:59 PM Pacific Time, plus one Fable limit reset. Max and Ultra subscribers get one to three Fable resets, expiring 8 March 2027 at 11:59 PM Pacific. After the launch week, Fable usage is billed at API rates through usage credits.

The safeguards have also improved, by Anthropic’s own measure: Fable 5.2’s safeguards blocked about 55 percent fewer requests than Fable 5.1’s in early testing — meaning fewer legitimate requests are caught by the safety filters. That matters for developers, because a blocked request is a failed task, not just an inconvenience.

The IPO in the background

The launch arrives as Anthropic prepares to go public. The company reported $16.8 billion in revenue in the third quarter of 2026, up 45 percent, ahead of a planned IPO — a figure covered on caawnews last month. CNBC reports that the IPO could come as soon as next week, at a valuation of $1 trillion to $1.5 trillion, down from a floated $2 trillion. That valuation is reported, not confirmed, and Renaissance Capital estimated on 26 October that the roadshow could begin the week of 9 November.

The timing is not a coincidence. A cheaper, faster flagship model, released days before a possible listing, is a statement to investors as much as to developers: Anthropic is competing on price and performance at the same time. The outside evaluator METR estimated on 5 November that Anthropic is improving its models on well-specified tasks at about twice its earlier pace.

For users, the practical question is simpler: whether the new model is worth switching to. For most everyday work, Anthropic itself says Opus 5.5 remains the default. For heavy coding, long documents and multi-step agent work, Fable 5.2 is now the cheaper and faster option — with the caveat that its best results come with heavy tool use, and heavy tool use costs money.

Frequently asked questions

What is the difference between Claude Fable 5.2 and Mythos 5.2? They are the same underlying model. Fable 5.2 is for general users and developers, with safeguards that block most high-risk cybersecurity and research-biology work. Mythos 5.2 has fewer restrictions and is available only to organisations verified through Anthropic’s Cyber or Life Sciences Verification Programs.

How much does Claude Fable 5.2 cost? $8 per million input tokens and $40 per million output tokens on the API — 20 percent less than Fable 5.1. An optional fast mode costs $16 and $80 and generates tokens up to 2.5 times faster.

What is the context window? 1 million tokens, with up to 128,000 tokens of output in a single response. The knowledge cutoff is June 2026.

Is Fable 5.2 available on the Free plan? No. It is available on all paid Claude plans — Pro, Max, Ultra, Team and Enterprise — and on the API. Pro and standard Team seats get it included up to 50 percent of weekly limits through 16 November 2026, plus one limit reset.

Is Fable 5.2 better than Opus 5.5? On Anthropic’s FrontierCode 2.0 coding benchmark it scored 49.1 percent against 45.5 percent for Opus 5.5, and it leads one independent ranking of 21 models. But Anthropic still recommends Opus 5.5 as the everyday default, and Fable’s best results come with heavy tool use that adds cost.

When will Anthropic go public? CNBC reports the IPO could come as soon as next week at a $1 trillion to $1.5 trillion valuation. That is reported, not confirmed by the company.

Key takeaways

• Anthropic released Claude Fable 5.2 and Mythos 5.2 on 9 November 2026 — one model, two safeguard levels.
• Price is $8/$40 per million input/output tokens, a 20% cut from Fable 5.1; fast mode is $16/$80, up to 2.5× faster.
• Specs: 1M-token context, 128K max output, June 2026 cutoff, available on the Claude API, AWS Bedrock, Google Cloud and Microsoft Foundry.
• Benchmarks: 49.1% on FrontierCode 2.0 (Anthropic’s own); first of 21 models at 67.1% in Artificial Analysis VIALS-AA — but with 4.7× the tool calls of GPT-6 Astra.
• All paid Claude plans get Fable 5.2 (not Free); launch-week offer runs through 16 November 2026.
• IPO reported by CNBC at $1T–$1.5T, possibly as soon as next week — reported, not confirmed.