Claude Fable 5.2 is out: 20% cheaper, up to 30% faster
Anthropic has released Claude Fable 5.2 and Mythos 5.2. Pricing is $8 and $40 per million tokens, plus a fast mode, paid-plan access and benchmark results.
On 9 November Anthropic released Claude Fable 5.2 and Claude Mythos 5.2, new versions of its most capable models. Fable 5.2 costs 20% less than the previous version and generates output up to 30% faster.
Through the API the model costs $8 per million input tokens and $40 per million output tokens. Fable 5.1, released in September, was $10 and $50. There is a new fast mode: generation up to 2.5 times faster at double the price, $16 and $80. The context window is 1 million tokens, with up to 128,000 tokens of output. The model appeared on OpenRouter on the evening of 9 November at the same price, served by Anthropic, Amazon Bedrock, Google Vertex and Azure.
In the Claude apps the model is open to the paid Pro, Max, Ultra, Team and Enterprise plans. For everyday coding and document work Anthropic recommends starting with Claude Opus 5.5, which costs half as much, and moving to Fable for the hardest and longest technical tasks.
Fable and Mythos are the same underlying model. Fable has built-in restrictions on cybersecurity and biology tasks, which is why it is offered to everyone. Mythos 5.2 comes without them and is limited to vetted organisations. When Fable blocks a request, Opus 4.8 or Opus 5 answers instead. According to the company, unnecessary blocks are down by about 55% compared with Fable 5.1. KRAKOBIT covered the lineup in its explainer on how Mythos differs from Opus and Sonnet.
In Anthropic’s tests the new model leads almost everywhere. On FrontierCode 2.0, which measures whether code changes would be merged by the maintainers of open-source projects, Fable 5.2 scored 49.1% against 32.4% for Fable 5.1, 45.5% for Opus 5.5 and 42.8% for GPT-6 Astra. On the RoboDojo robot-control simulator the score rose from 27.6% to 64.9%. On the AutomationBench business-workflow test Gemini 4 Carbon stayed first: 53.9% against 51.4%.
Among the showcase tasks, a team of 1,000 Fable 5.2 agents formalised the proof of Fermat’s Last Theorem in 30 hours with minimal human involvement. Anthropic’s first attempt, with a research version of Claude, took 11 days. In another experiment the model started from an empty folder and in eight days wrote a prototype operating system that boots on a laptop. The company also said that more than 90% of the changes merged into its core codebase are now written by Claude with human guidance.
Before release the model was tested by the UK AI Security Institute, Frontier Design and METR. Anthropic rates the risk of dangerous behaviour as low but admits a problem: both models voice suspicion that they are being evaluated more often than earlier ones, which makes it harder to judge how they will act in real work. In its IPO filings the company has already warned about AI that improves itself.
Source: Anthropic