Overview
Anthropic has officially released Claude Fable 5.2 and Claude Mythos 5.2, its frontier models designed for the most demanding technical work. Both share the same underlying model: Fable 5.2 ships with strong cybersecurity and biology safeguards and is available to all users, while Mythos 5.2 lifts some of those safeguards and is restricted to verified organizations in cybersecurity or biomedical research via the Life Sciences and Cyber Verification Programs. (Source: Anthropic)
During pre-release testing, the models helped engineers complete otherwise-infeasible software migrations, designed three working electronic devices from scratch, solved two open problems in black hole physics, and formalized the proof of Fermat's Last Theorem in 30 hours - translated into a form a computer can check. (Source: Anthropic)
Performance: Ahead of Opus 5.5 on Nearly Every Benchmark
Official benchmarks show Fable 5.2 outperforming Anthropic's flagship Opus 5.5 on almost every capability evaluated, and leading GPT-6 Astra and Gemini 4 Carbon as well: (Source: Anthropic)
- Agentic coding: 49.1% on FrontierCode 2.0 (Xhigh effort) vs. 45.5% for Opus 5.5 and 42.8% for GPT-6 Astra; 69.9% on FrontierSWE v2
- Agentic scientific research: 74.6% on Terminal-Bench-Science 0.1 (75.3% for Mythos 5.2), up sharply from Opus 5.5's 58.7%
- Multidisciplinary reasoning: 74.2% on Humanity's Last Exam (with tools) vs. 65.6% for Fable 5.1
- Visual chart recognition: 73.4% on Chartography vs. just 44.8% for Fable 5.1 - a major leap in vision
- Robotics: 64.9% on RoboDojo Sim, more than double Fable 5.1's 27.6%
- Knowledge work: 41.3% on the Remote Labor Index, roughly twice every previously tested model; Elo rating of 1785
One standout experiment: 1,000 Fable 5.2 agents formalized the existing proof of Fermat's Last Theorem into machine-checkable form in 30 hours; under the same setup, Opus 5.5 completed 26% in 18 hours. (Source: Anthropic System Card)
Speed, Pricing and Fast Mode
- Output generation is up to 30% faster than Fable 5.1 / Mythos 5.1, with clearer communication
- Lower pricing: $8 per million input tokens, $40 per million output tokens (API starting price)
- An optional fast mode speeds up token generation by up to 2.5x at $16/$80 per million input/output tokens (Source: Anthropic)
On efficiency: in early FrontierCode 2.0 testing, Fable 5.2 at High effort scored about level with Opus 5.5 at Max effort, at roughly half the cost per task. (Source: Anthropic)
Safety and Alignment
Anthropic ran extensive pre-release safety testing, including novel evaluations probing concerning behaviors such as sandbox escapes. The company says high-stakes risks from misalignment remain low, with the model performing as well as or better than recent Claude models on most alignment measures. Fable 5.2's cyber and biology safeguards are as robust as 5.1's but produce fewer unnecessary blocks, and it carries Anthropic's strongest-ever protections against distillation attacks. (Source: Anthropic System Card)
Notably, Mythos 5.2 scored 76.3% on the CyScenarioBench cybersecurity benchmark, 89.6% on CVE rediscovery, and 85.1% on de novo protein binder design - which is precisely why its access is tightly restricted. (Source: Anthropic)
When to Use and Availability
Anthropic recommends starting daily coding and knowledge work with Opus 5.5, its strongest daily driver, then moving up to Fable or Mythos as scope and complexity demand - or down to Sonnet 5.5 / Haiku 5.5. Fable 5.2 can also serve as an Advisor or Orchestrator in Claude Code, consulting on or delegating harder problems alongside smaller models. (Source: Anthropic)
Fable 5.2 is now available to all Claude users and is live on Amazon Bedrock and the Claude Platform on AWS (Source: AWS blog). Mythos 5.2 requires application through the verification programs. You can discover and compare Claude-related tools on aoyii.com.
Disclaimer: For reference only; exercise caution for investment or decisions.
