Anthropic Unveils Claude Fable 5.2 and Mythos 5.2, Claiming Top Performance at Lower Cost

Anthropic has announced the launch of Claude Fable 5.2 and Claude Mythos 5.2, which it describes as its most advanced AI models to date. According to the company, the new models deliver superior performance across several benchmarks compared to OpenAI's GPT-6 Astra, while also being 20% less expensive at standard API rates.

Priced at $8 per million input tokens and $40 per million output tokens, the new models are cheaper than their predecessors, which cost $10 and $50 respectively. Cached input tokens are priced at $0.20 per million, an 80% reduction compared to GPT-6 Astra's $1 per million. Anthropic also claims the models can generate responses up to 30% faster than earlier versions.

Both models share the same underlying technology, but Fable 5.2 includes additional cybersecurity and biological safety restrictions, making it more widely available. Mythos 5.2 is targeted at approved organizations in cybersecurity and life sciences. The company reports that the new models outperform its previous Claude Opus 5.5 across most tested capabilities and beat GPT-6 Astra on several benchmarks.

On the FrontierCode 2.0 benchmark, Fable 5.2 scored 49.1%, compared to 45.5% for Claude Opus 5.5 and 42.8% for GPT-6 Astra. On Humanity’s Last Exam with tools, Fable 5.2 achieved 74.2%, ahead of Opus 5.5's 67.7% and GPT-6 Astra's 57.2%. Anthropic says the models excel in coding, reasoning, scientific research, and image understanding.

The company highlights that Fable 5.2 can handle complex software development tasks, such as fixing bugs, migrating systems, and coordinating multiple AI agents. In one test, an earlier version of the model helped migrate a production service from Python to Rust, making its startup approximately 26 times faster.

While the price reduction and performance gains position Anthropic more competitively against OpenAI, the company has not explicitly linked the lower prices to competition with GPT-6 Astra. The launch underscores the intensifying race among AI developers to deliver more capable and cost-effective models.