Anthropic launches new Claude versions with performance upgrades and price cuts
On Monday, September 9, Anthropic unveiled updated versions of its AI models: Claude Fable 5.2 and Claude Mythos 5.2. These new systems are designed for complex technical tasks, offering improvements in programming, reasoning, and image analysis, while also delivering responses up to 30% faster than the previous 5.1 generation.
In addition to the speed boost, the company has slashed API prices by 20%. The standard rate is now $8 per million input tokens and $40 per million output tokens, down from the previous $10 and $50. There’s also an optional fast mode that generates tokens up to 2.5 times quicker but charges double the standard rate: $16 and $80, respectively.
Performance and test results
In tests released by Anthropic, Fable 5.2 scored 69.9% on FrontierSWE v2, which evaluates programming agents, surpassing the 56.3% score of model 5.1. On Chartography, which measures the ability to interpret graphs without external tools, the score jumped from 44.8% to 73.4%.
In a mathematical demonstration, around a thousand Fable 5.2 agents worked in parallel for 30 hours to formalize a proof of Fermat’s Last Theorem in Lean, a language that enables computer verification of each step. Despite these advancements, Anthropic still recommends using Claude Opus 5.5 for most programming and analysis tasks.
Differences between Fable and Mythos, and access
Both versions share the same base model. Fable includes additional filters for high-risk cybersecurity requests and biological research. It’s available through Anthropic’s API and eligible paid Claude plans.
Mythos 5.2, however, has restricted access limited to verified organizations, such as those participating in the company’s cybersecurity and life sciences programs. This restriction aims to ensure greater control over the model’s use in sensitive areas.
Anthropic also noted a challenge in security testing: the two new models indicated they are being evaluated more frequently than other recent systems, making it difficult to predict their behavior outside the lab environment.
Frequently Asked Questions
- What’s the difference between Claude Fable 5.2 and Claude Mythos 5.2? They share the same base, but Fable adds filters for cybersecurity and biology, while Mythos requires prior verification for direct access.
- How much do the new models cost on the API? The standard rate is $8 per million input tokens and $40 per million output tokens. In fast mode, the rates double to $16 and $80.
- Where is Claude Fable 5.2 available? The model can be accessed through Anthropic’s API, paid Claude plans, and partner cloud platforms like AWS, Google Cloud, and Microsoft Azure.
- What did Claude do with Fermat’s Last Theorem? Fable 5.2 agents converted a known demonstration into a step-by-step, computer-verifiable version.
The updates reflect Anthropic’s strategy to enhance the efficiency and security of its models while making the technology more accessible to businesses and researchers.








