Anthropic Releases Claude 5.2 Dual Models: API Unit Price Drops 20% While Max Thinking Multiplies Cost
Frontier AI models · AI assistants · AI coding
Anthropic introduced Claude Fable 5.2 and restricted-access Mythos 5.2, featuring identical architectures with distinct safety filters. Although API unit prices decreased by 20%, maximum thinking intensity significantly increases token consumption and doubles overall task cost.
Anthropic announced Claude Fable 5.2 and Claude Mythos 5.2 on November 9, 2026. Sharing the same base architecture, Fable includes cybersecurity and biological guardrails for public availability, while Mythos requires vetting for specialized research entities. API pricing is set at $8 per million input tokens and $40 per million output tokens, a 20% unit price reduction from Fable 5.1. However, benchmarks by Artificial Analysis reveal that running Fable 5.2 at maximum thinking intensity incurs a weighted average task cost of $26.5—approximately 2.6 times that of Opus 5.5 at max—whereas setting it to high reduces task cost by nearly half while maintaining top scores.
Key takeaways
- Fable 5.2 is publicly accessible with safety classifiers, whereas Mythos 5.2 requires verification for cyber and life sciences research.
- API pricing is $8 input and $40 output per million tokens, though maximum thinking intensity doubles total task expenses due to token volume.
- The model achieved 49.1% on FrontierCode 2.0 and reached 64.9% in the RoboDojo dual-arm robotic control benchmark.
- Anthropic stated that in internal experiments, 1,000 Fable 5.2 agents formalized Fermat's Last Theorem within 30 hours.
Source：數位時代 Read the original ↗
Prepared, checked and automatically published by AI; not individually reviewed by a human. Translation checked by AI.