Anthropic debuted Claude Fable 5.2 and Mythos 5.2 on Monday, positioning the models as its response to OpenAI's GPT-6 Astra, which launched in early September. Anthropic says the new models deliver higher performance than Astra at a lower price, a pricing advantage that could help it attract and retain customers in a market where professionals may switch models based on cost as long as performance remains adequate.

The launch comes as Anthropic seeks to reaccelerate revenue growth after a third-quarter slowdown disclosed in its latest S-1 filing with the SEC, the registration statement companies submit when preparing to list shares publicly. The releases could also be among the company's last major product launches before its initial public offering.

According to Anthropic, Fable 5.2 and Mythos 5.2 cost 20% less than GPT-6 Astra for standard input and output tasks. Anthropic charges $8 per million input tokens and $40 per million output tokens, compared with $10 and $50, respectively, for Astra.

Tokens are units of information processed by AI models. One token of English text is roughly equivalent to one and a half words. Cached tokens are portions of a prompt a model has already processed and stored, so requests that reuse the same material are billed at a reduced rate. For prompts containing large amounts of information, such as documents, the pricing difference is greater: Anthropic charges 20 cents per million cached tokens, compared with $1 per million cached tokens for GPT-6 Astra, making its models 80% cheaper for that use case.

Anthropic reported strong results from pre-release testing. The company said the models “helped our engineers complete otherwise-infeasible software migrations, designed three working electronic devices from scratch, solved two open problems in the physics of black holes, made substantial progress on robotics evaluations, and formalized the proof of Fermat's Last Theorem in 30 hours—among many other applications.”

The company also said Fable 5.2 and Mythos 5.2 outperformed competitors on coding evaluation benchmarks. More than 90% of Anthropic's own code is written by AI, according to the company. Anthropic's Claude Code initially gained an advantage among programmers, but OpenAI's Codex has gained ground in recent months because of its lower pricing, The Information reported in late October. Anthropic's combination of lower pricing and claimed performance improvements could help it regain those users if the results are confirmed in real-world testing.

Anthropic said the new models are designed for the “most ambitious, open-ended technical work,” putting them in competition with Astra at the frontier of AI capabilities. For routine tasks, however, the company recommends its more cost-effective Opus 5.5 model. Opus 5.5 costs more than OpenAI's direct competitor, GPT-6.1 Sol, although its value will depend on whether it produces high-quality results more consistently. Users will determine how the models compare as they begin using the latest releases in daily work.

Cybersecurity is another area of competition with OpenAI. Anthropic said an unnamed security company “found that one in six confirmed bugs found by Claude Mythos 5.2 were missed by all of the other frontier models they tested.” OpenAI debuted its cybersecurity-focused GPT-6 Cyber model late last week, according to Fortune.

Anthropic also said the new models solved two theoretical physics problems related to black holes and identified a way to accelerate research into gene therapies. The company is emphasizing the models' mathematical capabilities as OpenAI draws attention for solving mathematical proofs that have challenged academics for decades, including the 90-year-old Navier–Stokes Millennium Prize Problem.

A team of 1,000 agents powered by Fable 5.2 produced a formalization of a mathematical proof in 30 hours. Formalization translates a proof into a format that a computer can verify step by step. The result compared with 11 days for a previous, unreleased research version of Claude. Anthropic said faster formalization could help mathematicians verify proofs more quickly and “spend more time exploring novel ideas,” amid criticism from parts of the mathematics community over OpenAI's investment in solving proofs and its impact on the field.

Hardware and robotics are another area in which Anthropic appears to be placing greater emphasis than OpenAI. The company said the new models designed and produced three working electronic devices from a one-paragraph prompt, including circuit boards, 3D-printed enclosures, and software.

Anthropic also highlighted safety safeguards that block “high-risk cyber activities including penetration testing, exploit generation, and binary-based vulnerability, as well as most tasks in research biology.”

Claude Fable 5.2 is available immediately across platforms, including Amazon Web Services, Google Cloud, and Microsoft Azure, giving teams already running workloads on those clouds a direct path to test the models against their existing workflows. Access to Mythos 5.2 is available by application to cybersecurity professionals and life sciences researchers—the same fields where Anthropic highlighted the models' most distinctive results. Claude Enterprise customers can access Mythos 5.2 through Anthropic's Claude Security product.

This story was originally featured on Fortune.com: https://fortune.com/2026/11/09/anthropic-debuts-updated-claude-fable-and-mythos-models-undercutting-price-of-openais-gpt-6-astra-by-up-to-80/