Anthropic Launches Fable 5.2 with Price Cuts and Cybersecurity Gains Ahead of IPO

Key points

• Anthropic launched Claude Fable 5.2 and Mythos 5.2 on Monday, with Fable 5.2 available on all major cloud platforms and Mythos 5.2 restricted to Cyber and Life Sciences verification programs.
• The new models generate output up to 30% faster than the 5.1 generation and cost 20% less, with Fable 5.2 priced at $8 per million input tokens and $40 per million output tokens.
• Fable 5.2 scored 49.1% on FrontierCode 2.0 and 69.9% on FrontierSWE v2, outperforming OpenAI’s GPT-6 Astra and Anthropic’s Opus 5.5 in coding benchmarks.
• On Irregular’s CyScenarioBench, Fable 5.2 achieved a 76.3% success rate, surpassing GPT-6 Cyber’s 61% in multi-stage cyberattack simulations.
• Anthropic is preparing for a potential IPO, targeting a valuation of $1 trillion to $1.5 trillion, down from a previous $2 trillion target.
• The models show fewer safety blocks, with a 55% reduction in overall blocks, but still restrict high-risk cybersecurity and biology activities.

Background

Anthropic previously released Fable 5.1 and Mythos 5.1, with Mythos restricted to vetted organizations due to its advanced cybersecurity capabilities. The company had called for slowing the pace of AI development in September, and this launch is the first frontier model release since then. Earlier concerns about AI steganography in Fable-5.1 highlight ongoing safety challenges, though Anthropic claims improved alignment in the 5.2 versions.

How outlets are covering it

The New Stack emphasizes the price cut and coding benchmark improvements, noting that Fable 5.2 is 20% cheaper than GPT-6 Astra. CNBC highlights the IPO context, with Anthropic aiming for a $1 trillion to $1.5 trillion valuation, and notes the tension between safety and commercial pressures. The Hacker News focuses on cybersecurity benchmarks, where Fable 5.2 outperforms GPT-6 Cyber in multi-stage attack simulations. The GitHub Blog details the integration of Fable 5.2 into GitHub Copilot, noting that it requires data retention by default for safety classifiers, with zero data retention available for eligible enterprises.

Why it matters

The launch of Fable 5.2 and Mythos 5.2 signals Anthropic’s push to compete with OpenAI and Google in high-end AI capabilities, particularly in coding and cybersecurity. The price cut and performance gains could make Fable 5.2 a more attractive option for complex tasks, while the IPO preparation underscores the company’s commercial ambitions. The cybersecurity improvements and safety measures are critical as AI models are increasingly used in sensitive environments, and the reduced safety blocks may reduce interruptions for legitimate work.

What to watch

Anthropic is expected to begin its IPO roadshow soon, with investors evaluating the company’s balance of safety and commercial growth. The rollout of Fable 5.2 in GitHub Copilot and other platforms will continue, with gradual access for users. Enterprise Frontier Safeguards will move into private beta, with early customers including Bridgewater, Salesforce, and Stripe. Anthropic will likely continue to refine its safety measures and expand the use of Fable 5.2 in scientific and technical applications.