Anthropic has officially released Claude Fable 5.2 and Claude Mythos 5.2, its next-generation frontier AI models engineered for demanding technical challenges across software engineering, mathematical formalization, robotics, and biomedical research. Claude Fable 5.2 is available immediately across the Claude Platform, Claude apps, AWS Bedrock, Google Cloud, and Microsoft Foundry, with standard API pricing starting at $8 per million input tokens.
Key Highlights and Generational Upgrades
- 30% Faster Output Generation: The 5.2 architecture delivers up to 30% faster standard output generation compared to Fable 5.1 and Mythos 5.1, alongside an optional Fast Mode accelerating output tokens by up to 2.5x.
- 20% Developer API Price Cut: Standard developer API pricing drops to $8 per million input tokens and $40 per million output tokens, down from the previous $10 and $50 baseline.
- Frontier Coding Dominance: Claude Fable 5.2 set the highest score of any released model on FrontierCode 2.0 at 49.1% (Xhigh effort), matching Opus 5.5 Max effort at High effort for roughly half the compute cost per task.
- Autonomous Mathematics and Capsid Design: A coordinated swarm of 1,000 Fable 5.2 agents fully formalized Fermat’s Last Theorem in 30 hours, while Mythos 5.2 accurately modeled 28 of 34 complex gene-therapy protein capsids.
- Public Safeguards vs. Defensive Verification Split: While Fable 5.2 ships with robust public safeguards against cyber and biological misuse, Mythos 5.2 unlocks the identical core weights for vetted institutions under Anthropic’s Life Sciences and Cyber Verification Programs.
Claude Fable 5.2 and Mythos 5.2: Model Specifications and API Pricing
| Metric / Specification | Claude Fable 5.2 | Claude Mythos 5.2 | Claude Opus 5.5 (Reference) |
|---|---|---|---|
| Core Architecture | Shared 5.2 Frontier Weights | Shared 5.2 Frontier Weights | Claude 5.5 Hybrid Transformer |
| Context Window | 1,000,000 Tokens (1M) | 1,000,000 Tokens (1M) | 1,000,000 Tokens (1M) |
| Maximum Output Tokens | 128,000 Tokens (128K) | 128,000 Tokens (128K) | 128,000 Tokens (128K) |
| Standard Input Pricing | $8.00 per million tokens | $8.00 per million tokens (Vetted) | $15.00 per million tokens |
| Standard Output Pricing | $40.00 per million tokens | $40.00 per million tokens (Vetted) | $75.00 per million tokens |
| Fast Mode Token Pricing | $16.00 Input / $80.00 Output | $16.00 Input / $80.00 Output | Unavailable |
| Safety and Safeguard Tier | Public Biological and Cyber Safeguards | Vetted Defensive Research Programs | Standard Public Safety Guardrails |
| Cloud Infrastructure | Claude Platform, AWS, GCP, Azure | Life Sciences & Cyber Programs | Claude API, Web, Major Clouds |
| Default Client Effort | Medium Effort (Apps & Claude Code) | Custom Defensive Workflows | Standard Adaptive Reasoning |
Frontier Benchmark Performance and Availability
Anthropic evaluated Claude Fable 5.2 across rigorous frontier benchmarks, demonstrating decisive gains over both past iterations and competing frontier models such as OpenAI’s GPT-6 Astra and Google Gemini 4 Carbon.
In agentic software engineering on FrontierCode 2.0, Fable 5.2 at High effort scored 45.3%, effectively tying Opus 5.5 at Max effort while consuming approximately half the token cost per task. When scaled to Xhigh effort, Fable 5.2 achieved a state-of-the-art score of 49.1%, leading all publicly tested frontier systems. The model also posted 69.9% on FrontierSWE v2, 83.5 on BenchCAD 2.1 hardware engineering at Max effort, and 82.1% on the WANDR agentic search benchmark.
Beyond code and simulation, Fable 5.2 demonstrated strong zero-shot physical dexterity on RoboDojo simulation benchmarks, scoring 64.9% in controlling simulated two-arm robotic arms, outperforming custom robotics AI systems without task-specific pre-training. In formal mathematics, a distributed cluster of 1,000 Fable 5.2 agents completed the machine-verifiable step-by-step translation of Fermat’s Last Theorem in 30 hours, a milestone where Opus 5.5 reached only 26% completion before plateauing.
Claude Fable 5.2 is accessible today for all paid Claude subscribers across Claude.ai, Claude Code (defaulting to Medium effort), and major enterprise cloud platforms including Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Azure Foundry.
PFD Context
Anthropic’s architectural split between Fable 5.2 and Mythos 5.2 marks a deliberate strategy to balance frontier capability with biosecurity oversight. By dropping developer API token pricing by 20% while positioning Claude Opus 5.5 as the recommended daily driver, Anthropic establishes Fable 5.2 as an industrial-grade reasoning engine reserved for complex agent swarms, hardware CAD modeling, and long-horizon autonomous tasks.
