Anthropic Unveils Claude Sonnet 5, Resumes Fable and Mythos Models
Anthropic has officially launched Claude Sonnet 5 and restored access to its Fable and Mythos frontier models after resolving regulatory and technical hurdles. The company paused operations for 18 days following a U.S. government export control directive, which targeted risks tied to safety vulnerabilities in its highest-capability systems. The pause ended after Anthropic deployed an updated automated classifier to address a bypass method discovered by Amazon researchers, which allowed Fable 5 to identify software vulnerabilities and generate exploitation code.
Regulatory Challenges and Safety Enhancements
The export control mandate, enacted on June 12, forced a temporary blackout on global access due to gaps in real-time nationality verification. During the shutdown, security evaluations confirmed the vulnerability wasn’t unique to Fable 5—older models from competitors like GPT-5.5 and Kimi K2.7 also exhibited similar behaviors. To comply with federal directives, Anthropic engineered a safety classifier that blocks ambiguous prompts with a statistical probability of malicious intent. Internal tests show the system prevents the reported exploitation technique in over 99% of cases, though it now flags more benign requests during routine development.
Performance and Cost Efficiency
Claude Sonnet 5 outperforms its predecessors in key benchmarks, achieving 63.2% on SWE-bench Pro and 80.4% on Terminal-Bench 2.1, while maintaining lower costs than Opus 4.8. The model’s introductory pricing of $2.00 input and $10.00 output per million tokens through 2026 positions it as a cost-effective option for developers. Anthropic’s engineering teams are also shifting autonomous agents to Sonnet 5 to reduce operational expenses while preserving high execution capacity.
Real-World Applications
Organizations like Rakuten and Zapier are already integrating Sonnet 5 into production workflows. Rakuten’s teams used the model to automate code reviews, testing submissions and validating results before human engineers approve changes. Zapier incorporated the system into its core product, demonstrating its utility in streamlining software development pipelines. As Anthropic navigates regulatory scrutiny, the renewed focus on safety and efficiency underscores the evolving landscape for frontier AI systems.
Source: AI News. AI-assisted editorial synthesis — TechnoExpress.