Anthropic Fable 5 to Be Officially Released: 18-Day Export Control Ends, Restoring Global Access to AI Programming Model.

TradingKey - On June 30, Eastern Time, Anthropic announced on social platform X that the U.S. Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5, and official access to both models will resume on July 1. This brings a close to the 18-day ban that began on June 12.

[Source: X]

The turmoil began on June 12. Citing national security concerns, the U.S. Department of Commerce issued an export control directive prohibiting any foreign nationals (including Anthropic's own foreign employees) from accessing Fable 5 and Mythos 5. Following the directive, Anthropic was forced to shut down both models across the board immediately.

According to The Wall Street Journal, the immediate trigger for the restrictions was the discovery by Amazon's internal researchers that specific prompts could bypass Fable 5's safety guardrails, causing the model to output information that could be used for cyberattacks.

Over the past 18 days, the global developer community experienced severe "withdrawal symptoms." A large number of AI startups relying on Fable 5 to build products ground to a halt, independent developers accustomed to writing 10,000 lines of code a day were forced back into inefficient modes, and multinational teams paused their overseas business expansion due to tedious compliance screening approvals.

The turning point came last week. After Anthropic co-founder Tom Brown took over negotiations with officials, communication efficiency between the two sides improved significantly. On the evening of June 30, Commerce Secretary Howard Lutnick officially signed off on revoking the export controls. Lutnick stated in a letter to Anthropic that the company had agreed to "proactively detect and address model-related safety risks," cooperate with the government in developing release protocols for future models, and report any "恶意活动" (malicious activity) detected in the models.

Fable 5 is the Mythos-class flagship model officially released by Anthropic on June 9. The company claims it has reached industry-leading levels in software engineering, knowledge work, visual understanding, and scientific research.

Data also supports this claim. According to public evaluation data, in the automated programming benchmark SWE-Bench Pro, Fable 5 achieved a pass rate of 80.3%, far exceeding GPT-5.5's 58.6% and Gemini 3.1 Pro's 54.2%, ranking first in the Agent Arena benchmark. It is reported that Fable 5 can migrate and compress a 50-million-line Ruby codebase in under a day, a task that would take humans more than two months.

According to Anthropic's announcement, access to Fable 5 will be restored in phases, with specific plan availability and usage limits subject to official updates. Meanwhile, Anthropic stated that it will continue to coordinate with the U.S. government to expand the availability of Mythos 5 under the "Glasswing" program.

Just hours before the lifting of the ban was announced, Anthropic released the mid-tier model Claude Sonnet 5 overnight. Positioned as the "most capable Sonnet model for agents to date," it can autonomously formulate plans and utilize tools such as browsers and terminals to complete multi-step tasks independently. In terms of pricing, Sonnet 5 is priced significantly lower than Opus 4.8; based on standard pricing, the cost per million tokens is about 60% lower than Opus 4.8. Analysts suggest this is a precise strategic layout to capture user mindshare with a highly cost-effective product prior to Fable 5's return.

In some ways, the 18-day ban turned out to be an accidental "hunger marketing" campaign. Now that the ban has been lifted, the enthusiasm of global developers has been reignited. Fable 5's official return tomorrow will serve as a dual test of Anthropic's technological prowess and its government relations capabilities.