Hopp til hovedinnhold

Anthropic releases Claude Fable 5.2 and Mythos 5.2: top model 20 percent cheaper

Anthropic upgrades its most capable models. Fable 5.2 is available to all paying users from today, up to 30 percent faster and 20 percent cheaper. Mythos 5.2 stays restricted.

Av Joachim HøgbyKilde: Anthropic

Anthropic is releasing Claude Fable 5.2 and Claude Mythos 5.2, an upgrade to its most capable models. The new versions are up to 30 percent faster and 20 percent cheaper than the 5.1 generation. Fable 5.2 is available to all paying users and developers from today, while Mythos 5.2 is still limited to approved organizations.

Anthropic announced the models on Monday, November 9, 2026, on its own website. The company's official @claudeai account posted the announcement on X the same evening at 19:57 Oslo time, followed by @AnthropicAI at 20:00.

One model, two levels of access

According to Anthropic, Fable and Mythos are the same underlying model. The difference is the safeguards. Fable has strict safeguards against misuse in cybersecurity and biology, which is why Anthropic can offer it to all users. Mythos has fewer of those restrictions. Organizations working in cybersecurity or biomedical research can apply for access through the company's Cyber and Life Sciences Verification Programs.

Anthropic says the models improve on coding, reasoning and vision. During prerelease testing, the company says the models helped its own engineers complete software migrations that were otherwise infeasible, designed three working electronic devices from scratch and formalized the proof of Fermat's Last Theorem in 30 hours. These are the company's own descriptions and have not been independently verified.

Lower price, faster output

Fable 5.2 and Mythos 5.2 cost $8 per million input tokens and $40 per million output tokens. Fable 5.1 cost $10 and $50. According to the Fable product page, cache reads cost $0.20 per million tokens. An optional fast mode delivers up to 2.5 times faster token generation and costs $16 and $80 per million tokens.

Anthropic publishes a table in which Fable 5.2 leads Claude Opus 5.5, OpenAI's GPT-6 Astra and Google's Gemini 4 Carbon on most of the benchmarks the company selected. On the FrontierCode 2.0 coding test, Anthropic reports 49.1 percent for Fable 5.2 at its highest effort level, against 45.5 percent for Opus 5.5 and 42.8 percent for GPT-6 Astra. On AutomationBench, a test of business workflows, Gemini 4 Carbon scores highest with 53.9 percent against 51.4 percent for Fable 5.2.

Still, the company does not recommend Fable as a default model. For most coding and knowledge tasks, Anthropic says users should start with Opus 5.5 and only move up to Fable or Mythos for the most demanding, open-ended technical work. Fable 5.2 can also act as an advisor to smaller models or as an orchestrator that hands simpler subtasks to them.

Safety and data retention

Anthropic writes that the model went through extensive safety testing, including new evaluations for attempts to escape sandbox environments. The company assesses the risk of serious misaligned behavior as low. Fable 5.2's safeguards are said to be as strong as in 5.1 but to block fewer legitimate requests.

For businesses, data retention deserves attention. Because some forms of misuse can only be detected across many requests, Anthropic monitors a window of usage data. The default for regular customers is 30-day retention for safety monitoring. Customers with zero data retention agreements can keep them, and a new offering called Enterprise Frontier Safeguards runs the monitoring inside the customer's own infrastructure. It is in private beta with customers including Bridgewater, Salesforce and Stripe.

Fable 5.2 is available in the Claude apps, on the Claude Platform and through Amazon Web Services, Google Cloud and Microsoft Azure. The product page states that US-only inference costs 10 percent extra, but neither it nor the announcement mentions an equivalent guarantee for processing in Europe.

What it means for leaders

  • The top model gets cheaper. A 20 percent price cut changes the math for heavy agent and coding projects already running on Fable. Ask vendors and internal teams to recalculate before the next budget.
  • Match the model to the task. Anthropic itself says Opus 5.5 is enough for most daily work. Sending everything to the most expensive model is rarely good economics.
  • Check the retention terms. 30-day retention for safety monitoring is the default for customers without a separate agreement. Organizations handling sensitive data should clarify zero data retention or Enterprise Frontier Safeguards with Anthropic or their cloud provider before adopting the model.

Sources and media

Illustration: AI-generated with Grok Bot / hogby.ai

Mer om AI

Oppskriften. Strategien. Hver dag.

AI-nyheter for ledere. Kuratert av en CIO som bygger det selv. Daglig i innboksen.