Amara Osei

Anthropic launched Claude Fable 5.2 and Mythos 5.2 on Nov. 9, 2026, setting standard prices 20% below OpenAI’s GPT-6 Astra and cached-token prices 80% below the rival model. The launch answered GPT-6 Astra, released in early September 2026.

What does Anthropic’s pricing strategy change?\n

• Standard input: Anthropic listed Fable 5.2 and Mythos 5.2 at $8 per input token, against $10 for GPT-6 Astra.
• Standard output: Anthropic listed both models at $40 per output token, against $50 for Astra.
• Cached input: Anthropic charged $0.20 per million cached tokens (reused prompt data), versus $1.00 for Astra.

The 20% standard discount and 80% cached discount give Anthropic a direct argument for retaining price-sensitive professionals, provided performance holds. The release could be one of Anthropic’s last big launches before an IPO, following a third-quarter revenue slowdown disclosed in its Nov. 2 SEC S-1 filing (an IPO registration statement).

Can Fable and Mythos regain coding users?

Anthropic said Fable 5.2 and Mythos 5.2 outperformed competing systems on coding evaluation benchmarks. AI writes more than 90% of Anthropic’s own code, making benchmark strength commercially central to the release.

In pre-release testing, Anthropic 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 results came from Anthropic’s own testing.

Anthropic built an early lead among programmers, but The Information reported in late October 2026 that OpenAI’s Codex had gained ground because of its lower pricing. If Anthropic’s benchmark and real-world performance claims hold, the combination of lower prices and stronger results could win programmers back.

Which model should customers choose?\n

Anthropic positioned Fable 5.2 and Mythos 5.2 for its most ambitious, open-ended technical work rather than routine tasks. For daily work, it recommended the more cost-efficient Opus 5.5, although that model costs more than OpenAI’s GPT-6.1