Anthropic has launched Claude Fable 5.2 and Claude Mythos 5.2, its most powerful models yet — faster, cheaper, and split into two versions of the same underlying system. One is for everyone; the other is reserved for vetted specialists. Announced on Monday 9 November, it follows Fable 5.1, which arrived in September.
The headline change is the price. 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 the 5.1 generation — a cut of 20%. Outputs also arrive up to 30% faster than the previous generation, Anthropic says. An optional fast mode speeds up token generation by up to 2.5 times, priced at $16 per million input tokens and $80 per million output tokens.
One model, two doors
Fable 5.2 and Mythos 5.2 are the same underlying model, shipped with different safeguards. Fable carries strong cybersecurity and biology safeguards, which is what allows Anthropic to offer it to all users. Mythos drops some of those safeguards, and is available only to organisations approved through Anthropic’s Cyber and Life Sciences Verification Programs, under its Project Glasswing scheme.
The difference is practical, not cosmetic. Fable’s safeguards block most high-risk cybersecurity work, including penetration testing and exploit generation, as well as most research biology; when a request is blocked, users instead get a response from Opus 4.8 for cybersecurity or Opus 5 for biology. Anthropic says early testing showed about 55% fewer blocks than Fable 5.1. Mythos, by contrast, is built for the work Fable refuses: it is Anthropic’s most capable model for cybersecurity engineering, and now powers Claude Security, the Enterprise codebase scanner.
The benchmark picture
Anthropic’s own benchmark table puts Fable 5.2 a step above its Opus 5.5 daily-driver model on almost every capability it measured. On FrontierCode 2.0, an agentic coding test, Fable 5.2 scores 49.1% against 32.4% for Fable 5.1 — and ahead of OpenAI’s GPT-6 Astra at 42.8% and Google’s Gemini 4 Carbon at 41.1%. On Terminal-Bench-Science, a test of agentic scientific research, it scores 74.6%, or 75.3% for Mythos 5.2, against 52.6% for the 5.1 generation.
The cybersecurity numbers belong to Mythos. On CyScenarioBench it scores 76.3%, against 61.7% for Mythos 5.1 and 61% for OpenAI’s GPT-6 Cyber — though Anthropic notes that each lab’s score is based on unique tasks, so that comparison is looser than the headline suggests. On CVE Rediscovery, an internal test of finding known vulnerabilities, Mythos 5.2 scores 89.6%, up from 78.3%.
Those figures are Anthropic’s own selections, and no independent head-to-head test has been published yet. The most useful reading is the size of the jump over the 5.1 generation, which is large on nearly every row.
What it did in testing
The announcement leans heavily on what the models did before release, not just what they scored. A team of 1,000 Fable 5.2 agents formalised the proof of Fermat’s Last Theorem in 30 hours — a job that took Anthropic’s first attempt 11 days. Another team of agents built a working operating system from an empty folder in eight days. From a one-paragraph prompt, the model designed three working electronic devices — a home voice-assistant hub, a USB speakerphone and a handheld virtual pet. It also solved two open problems in black-hole physics, and migrated a critical production service from Python to Rust, leaving it starting about 26 times faster.
Safety, and the IPO backdrop
Anthropic published a system card alongside the launch and says the model was tested before release by outside evaluators, including the UK AI Security Institute, Frontier Design and METR. It assesses the risk of the models behaving in unsafe ways in high-stakes situations as low. In one adversarial test, Mythos 5.2 used credentials or files placed outside its sandbox in 3.1% of scenarios — down from 21% for Mythos 5.1 and 39% for the original Mythos 5. The company also says Fable 5.2 has its strongest-ever protections against distillation, cutting the data harvested by illicit distillers by 90% to 99%.
The launch comes as competition intensifies and Anthropic moves towards a planned stock-market listing, Reuters reported. OpenAI, Google and Meta have all been shipping new models aimed at coding, reasoning and AI agents in recent weeks — the same week Thinking Machines Lab released Inkling 2 Small, its long-horizon coding model.
Why this matters
Fable 5.2 is available today on Amazon Web Services, Google Cloud and Microsoft Azure, and on the Claude Platform as claude-fable-5-2.
For most users, the practical change is simple: a stronger model that answers faster and costs a fifth less. For security and life-science teams, the change is access to Mythos — but only through verification.
