Anthropic launches Claude Sonnet 5 as US clears Fable 5 for return

Anthropic released Claude Sonnet 5 on Tuesday morning, expanding its context window to 1 million tokens, while simultaneously securing clearance from the US Department of Commerce to restore access to its sidelined Fable 5 and Mythos 5 models. The dual milestones resolve weeks of regulatory negotiations between the AI developer and the Trump administration over model safety safeguards.

The Department of Commerce lifted export controls on Claude Fable 5 and Mythos 5 following weeks of negotiations, allowing Anthropic to begin restoring access on Wednesday. To avoid similar regulatory blocks with Sonnet 5, Anthropic limited the new model's cyber capabilities, keeping its safety profile in line with older, less restricted models.

Technical specifications of Sonnet 5

According to developer documentation, Claude Sonnet 5 introduces significant API changes. The model features a 1 million token context window and a maximum output limit of 128,000 tokens, with adaptive thinking enabled by default.

Willison's analysis notes several key changes to the API structure:

• Sampling parameters like temperature, top_p, and top_k are no longer supported.
• Adaptive thinking is turned on by default unless explicitly disabled in the API payload.
• The model retains the same set of tools and platform features as Claude Sonnet 4.6.

Pricing and promotional rates

Anthropic positioned Sonnet 5 as offering performance close to its Opus 4.8 model but at a lower price point. The company kept pricing flat against Sonnet 4.6 at $3 per million input tokens and $15 per million output tokens.

To encourage adoption, Anthropic is running an introductory discount through August 31, reducing rates to $2 per million input tokens and $10 per million output tokens. The pricing structure matches the previous generation, ensuring developers do not face immediate cost increases when upgrading. The promotional discount is aimed at developers migrating large-scale pipelines to the new 1-million-token context window.

Regulatory clearance and Fable 5's return

A reporting on the regulatory decision notes that the US Department of Commerce lifted export controls on Claude Fable 5 and Mythos 5 on Tuesday. Anthropic had sidelined Fable 5 in early June following a Friday evening ultimatum from the Trump administration, which raised concerns over potential jailbreaks of the underlying technology.

Fable 5 is Anthropic's consumer-facing model built on the same core technology as Mythos 5, but packaged with additional safety safeguards. In a statement posted to X, Anthropic confirmed the policy resolution and its plans to redeploy the models.

"We’ve received notice that the Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5. We’ll begin restoring access tomorrow, and will share an update soon. We’re grateful to our users for their patience, and to everyone who worked with us on redeploying the models."

Cyber capabilities and safety guardrails

To prevent Sonnet 5 from facing similar export restrictions, Anthropic designed the model with reduced cyber capabilities. Based on our analysis and testing, Sonnet 5 is significantly less capable at cyber tasks than Mythos 5, allowing it to clear the safety thresholds previously established for older models.

The system card states that Sonnet 5's safeguards align with those applied to Opus 4.7 and Opus 4.8. By keeping cyber capabilities below the government's intervention threshold, Anthropic successfully launched the model globally without triggering the export controls that had previously halted its Mythos-class deployments.