Scoop: Anthropic's Claude Fable 5 returns to public access as government vetting moves toward a formal process

Anthropic's most powerful public model, Claude Fable 5, is on track to return to users after a 15-day government security freeze. The Trump administration is close to allowing access restoration, marking a thaw in a four-month standoff between Anthropic and the government. For AI builders, this signals that formal government vetting of frontier models is moving from concept to operational reality.

What happened

Anthropic's Claude Fable 5 was taken offline due to security fears by the government, just three days after its June 9 launch. The model had been promoted as the most capable ever released to the public and was briefly available at no extra cost on several paid Claude subscription plans through June 22 before access vanished. During that short window, users praised Fable 5 for deep thinking and sophisticated coding, with the newsletter "Vibe Check" calling it "the best coding model in the world." Now, sources tell Axios that both Lutnick and Treasury Secretary Scott Bessent have helped defuse the fight, and restoration of access is imminent. Both Anthropic and OpenAI are pushing the administration to codify a process for reviewing new models, as envisioned in President Trump's June 2 executive order.

Why AI builders should care

Fable 5's blackout was unprecedented: a top-tier model pulled from users' hands by government intervention. For developers, this creates a new risk factor when building on frontier models. Government vetting is moving from informal pressure to a potential formal framework that could affect how powerful models are deployed, who gets access, and under what conditions. Fable 5 has also become a benchmark: every new model, especially open-source ones, is being measured against its capabilities. Understanding the vetting trajectory helps builders decide where to invest integration effort and when to diversify model dependencies.

Practical implications

The immediate practical detail for builders is pricing: $10 per million input tokens and $50 per million output tokens, with a 90% input token discount for prompt caching. That is less than half the price of the earlier Claude Mythos Preview, making Fable 5 a cost-competitive option for high-throughput coding and reasoning tasks when it returns. Teams should also prepare for potential access restrictions based on customer segment or region, as the government's review criteria are still being negotiated. Developers who tested Fable 5 during its brief availability saw dramatic gains in code quality and reasoning depth, suggesting it will remain a strong candidate for agentic workflows and complex problem-solving.

Caveats

The return timeline is based on a single news scoop and may shift as government vetting and company policies evolve. Details of the vetting process, eligibility for access, and any post-vetting deployment constraints remain uncertain. Different customer segments or geographic regions may face different access terms. Builders should plan for possible intermittent access as the framework solidifies.

This development marks a practical path forward for balancing security concerns with rapid AI experimentation, but the details will matter for production deployments.