Key Notes
- Anthropic launched Fable 5.2 and Mythos 5.2, with broad Fable access and verified access for Mythos.
- Standard input and output token rates are 20% below version 5.1, alongside claimed speed and capability gains.
- The release expands Anthropic’s enterprise pitch ahead of its proposed Nasdaq IPO.
Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 on November 9, pairing its latest push into demanding technical work with lower prices. The company says the models improve coding, scientific research and financial analysis, while generating output up to 30% faster than their 5.1 predecessors.
Fable is broadly available; direct Mythos access requires verification for cybersecurity or life-sciences work. They share the same underlying model, with stronger safeguards on Fable limiting potentially harmful uses.
Lower Prices Target More Demanding Work
Standard pricing is $8 per million input tokens and $40 per million output tokens, according to Anthropic’s model documentation. Both rates are 20% below the $10 and $50 listed for 5.1. At those base rates, a workload consuming one million input tokens and one million output tokens would cost $48, against $60 previously.
An optional fast mode costs twice as much, at $16 for input and $80 for output per million tokens. Anthropic says it can generate tokens up to 2.5 times faster. That is a separate speed claim from the standard model’s improvement over version 5.1.
The developer documentation positions Fable for sessions that run for hours, multistep research and finished analytical work. Even after the cut, its standard token rates remain twice those of Opus 5.5. Buyers therefore have a reason to reserve the more expensive model for jobs where better results justify the premium.
Financial Research Joins the Sales Pitch
Early feedback in the launch announcement includes quantitative traders and investment researchers testing Fable on research coordination, financial documents and ambiguous analytical questions. Those accounts describe trial workflows, rather than measured investment returns.
For financial firms, a lower token bill is only part of the calculation. Time spent checking a model’s work, correcting calculations and rerunning failed tasks also determines whether automation saves money.
Black-Hole Findings Appear in Preprints
Anthropic says Fable 5.2 solved two open problems in black-hole physics. The supporting preprints examine a phase transition and a mathematical bootstrap limit. Their provenance statements describe AI-generated results checked by physicists; the second also credits Opus 5.5 with independently reaching the same result.
The Launch Arrives Before a Proposed IPO
Anthropic is bringing that product pitch to prospective shareholders as it pursues a Nasdaq listing under the proposed symbol ANT. Its November 2 amended prospectus reported third-quarter revenue of $16.956 billion and a $567 million operating loss. Adjusted operating income was $2.463 billion, a separate measure that excludes certain expenses.
The quarterly figures also came with approximately $895.4 billion of purchase and lease commitments, substantially all tied to computing and infrastructure. These are multiyear obligations extending into 2045, rather than an immediate bill.
For investors weighing the proposed offering, cheaper access to a stronger model creates a concrete commercial challenge: Anthropic needs more usage, more valuable work or lower delivery costs to compensate for charging less per token.
