Cheaper pricing and a new Fast mode
Fable 5.2 and Mythos 5.2 cost $8 per million input tokens and $40 per million output tokens, down from $10 and $50 for Fable 5.1. That is 20 percent below OpenAI's GPT-6 Astra, which still charges $10 and $50 for standard prompts, according to The New Stack. The models also come with an optional Fast mode, which generates tokens up to 2.5 times faster and costs $16 per million input tokens and $80 per million output tokens. Fable 5.2 is available on the Claude API, Claude Code, Amazon Web Services, Google Cloud and Microsoft Azure. Mythos access remains limited to organisations in Anthropic's Cyber and Life Sciences verification programmes.
Same model, two versions
Fable 5.2 and Mythos 5.2 are the same underlying model, Anthropic said. Fable carries cybersecurity and biology safeguards and is open to all customers, while Mythos is available to vetted organisations working in those dual-use areas. Anthropic still recommends that most users start with Claude Opus 5.5, its strongest daily driver, and move up to Fable only for the most ambitious technical work. The models have a context window of 1 million tokens and can produce up to 128,000 output tokens.
Benchmarks: strong on coding, not a clean sweep
On Anthropic's own benchmark table, Fable 5.2 scored 49.1 percent on FrontierCode 2.0, ahead of Opus 5.5 at 45.5 percent, GPT-6 Astra at 42.8 percent, Gemini 4 Carbon at 41.1 percent and Fable 5.1 at 32.4 percent. It also led on FrontierSWE v2 at 69.9 percent, on the Remote Labor Index at 41.3 percent against 17.9 percent for Fable 5.1, and on RoboDojo Sim at 64.9 percent. These are Anthropic's own results, measured at each model's highest available effort, and not independent tests. The lead is not a clean sweep. Gemini 4 Carbon scored 53.9 percent on AutomationBench against Fable's 51.4 percent, and Opus 5.5 kept a small lead on the AA-Briefcase knowledge-work measure. Anthropic also said Fable 5.2 at High effort matched Opus 5.5's best FrontierCode score for roughly half the cost per task. Cursor, which added the model on launch day, reported 56.9 percent on its own CursorBench at default effort and 60 percent at maximum effort.
On launch day, Fable 5.2 also moved to the top of the Artificial Analysis Intelligence Index, taking the spot previously held by Gemini 4 Carbon, according to Mashable. The launch follows a rapid run of frontier model releases, including Google's Gemini 4 Carbon, Meta's Muse Wonder and OpenAI's GPT-6 Astra.
What it did in testing
Anthropic said that in prerelease testing, the models helped its engineers finish software migrations that had previously looked infeasible, designed three working electronic devices from scratch, solved two open problems in the physics of black holes and formalised the proof of Fermat's Last Theorem in 30 hours. The Fermat work used a team of 1,000 agents with minimal human intervention. An earlier research version of Claude needed 11 days for the same task, using a purpose-built harness and human guidance. In coding, Anthropic said more than 90 percent of the changes merged into its core codebase are now written by Claude with human guidance. A team of agents also migrated a critical production service from Python to Rust, and the service now starts about 26 times faster. In one experiment, an engineer asked Fable 5.2 to design a prototype operating system from an empty folder. The model coordinated more than 1,000 subagents over eight days and produced an OS that boots on a laptop, with its own kernel, drivers, file storage, windows, a login screen and an installer.
Safety and what developers must change
Anthropic said Fable 5.2's cyber and biology safeguards are as robust as those in Fable 5.1 but block legitimate work less often, with about 55 percent fewer blocks in early testing across Claude.ai, Claude Code and Cowork. The company also said the model has its strongest-ever protections against distillation, the illicit practice of training a model on another model's outputs. Its Enterprise Frontier Safeguards system, which watches for misuse across patterns of use, has entered private beta with customers including Bridgewater, Salesforce and Stripe. Developers moving from Fable 5.1 should note four breaking changes in Anthropic's migration guide: a prefix check on every account, thinking blocks that only 5.2 models can read, the retirement of the computer_20251124 tool, and new advisor pairing rules.
