Anthropic Launches Claude Sonnet 5, Unlocks Frontier Models After Federal Review
Anthropic has officially deployed Claude Sonnet 5 and restored access to its Fable and Mythos frontier models following an eighteen-day operational pause triggered by a US government export control directive. The models are now live across Anthropic's platform, cloud infrastructure, and partner networks.
The federal directive, enacted on June 12, forced the temporary suspension of Anthropic's highest-capability systems after researchers at Amazon documented a method to bypass the safety controls of Fable 5, causing the model to identify software vulnerabilities and supply exploitation code. The lack of real-time nationality verification systems at the time required a total access blackout for all global users.
The Security Fix That Cleared the Blockade
To resolve the federal mandate, Anthropic engineers trained an automated safety classifier targeting the specific bypass mechanism reported by Amazon. The new software layer functions with a wide safety margin, identifying and blocking ambiguous developer prompts that display a statistical probability of malicious intent. Internal validation data indicates the updated classifier prevents the reported exploitation technique in more than 99 percent of trials.
When a developer issues a prompt that triggers this boundary, the platform automatically routes the workload to the older Claude Opus 4.8 architecture to maintain continuity. The expanded safety margin introduces a trade-off: the automated system now flags benign requests more frequently during routine application development and software debugging.
Sonnet 5 By the Numbers
The immediate commercial focus targets the newly-deployed Claude Sonnet 5, and the benchmark data tells a clear story of generational improvement.
- SWE-bench Pro: Sonnet 5 scores 63.2% vs Sonnet 4.6 at 58.1% -- a 5.1-point jump in real-world software engineering tasks
- Terminal-Bench 2.1: Sonnet 5 hits 80.4%, up from 67.0% -- a massive 13.4-point leap in terminal-based agentic coding
- Pricing: Base input cost remains at $3.00/MTok and output at $15.00/MTok -- identical to Sonnet 4.6, meaning the 13-point Terminal-Bench improvement comes at zero additional cost
Engineering teams are already transitioning autonomous agents to Sonnet 5 to reduce operational expenditure while maintaining high execution capacity. Performance data validates that the system executes multi-step plans, operates terminal environments, and navigates web browsers without human intervention.
The Vulnerability That Wasn't Unique
Security evaluations conducted during the shutdown revealed a critical finding: the vulnerability identification behavior was not unique to Fable 5. Older and less capable architectures from multiple providers, including Claude Opus 4.8, GPT-5.5, and Kimi K2.7, duplicated the exact same results. This suggests the issue was not a model-specific flaw but a broader behavioral characteristic of frontier AI systems under adversarial prompting -- a finding with significant implications for how the industry approaches safety regulation.
What This Means for Developers
For developers building on Anthropic's stack, the Sonnet 5 release represents a no-compromise upgrade. The same pricing with substantially better agentic coding performance makes it the obvious migration target for production workflows. The restored Fable and Mythos models meanwhile signal that the US government is willing to lift restrictions when safety mitigations are demonstrably effective, setting a potential precedent for future regulatory interactions across the industry.
The key takeaway: Claude Sonnet 5 is shipping today at no premium, Fable and Mythos are back online with hardened safety classifiers, and the industry just learned that vulnerability exploitation in frontier models is a systemic challenge, not a bug limited to any single provider.
Comments