Claude 5.2 Released: Comprehensive Performance Upgrades, API Price Drops by 20% Anthropic released Claude Fable 5.2 and Mythos 5.2 at once. Both have identical underlying capabilities, differing only in their security measures: Fable is for general paid users, while Mythos is available to approved cybersecurity and life science institutions. Practical Capabilities: Programming: FrontierCode 2.0 score of 49.1%, surpassing Opus 5.5's 45.5%, GPT-6 Astra's 42.8%, and Gemini 4 Carbon's 41.1%. Robot Control: RoboDojo Sim score of 64.9%, Opus 5.5's 39.4%. Mathematical Research: 1000 AI agents collaborated, formalizing the existing proof of Fermat's Last Theorem into Lean code in just 30 hours, allowing for step-by-step verification by computers. Hardware Engineering: Starting with textual requirements, we independently completed component selection, four-layer PCB routing, 3D casing, and firmware design, achieving manufacturability design for three devices within one day. Biological Research: Mythos tested 34 viral capsid structures, with prediction errors controlled within 3 angstroms for 28 of them. Now let's look at the cost, which is the most interesting part: The API standard price has been reduced from $10/$50 to $8 per million token input and $40 per million token output, a direct price reduction of 20%. Generation speed has increased by up to 30%, and Fast Mode by up to 2.5 times, but the price has doubled to $16/$80. On FrontierCode 2.0, the Fable High score is 45.3%, almost matching the Opus 5.5 Max score of 45.5%, while the cost per task is only about half that of the latter. Specifications: 1 million token context window, 128,000 token maximum output, and adaptive thinking always enabled. For regular users, the Max and Ultra plans include Fable usage credits; Pro users also received a temporary credit discount from November 9th to 16th, after which additional usage will be charged separately. Free users are not currently eligible. Anthropic's usage recommendation: Use Opus 5.5 for daily tasks, and switch to Fable 5.2 for complex programming, scientific research, and tasks requiring long-term autonomous execution.