Item: Anthropic
Author: Karl Strauchman

Anthropic restored global access to Claude Fable 5 on July 1, 2026, one day after the U.S. Department of Commerce lifted an emergency export-control directive that had forced the model offline for 18 days. The company used the same morning to ship Claude Sonnet 5 at introductory pricing of $2 per million input tokens and $10 per million output, framing it as the cost-performance workhorse for agentic workloads.

The two announcements are being read as separate stories. They aren't. Fable 5 and Mythos 5 launched together on June 9 as a shared underlying model with different safeguard configurations. Three days later, Commerce cited national-security authorities and barred access by any foreign national, a restriction VentureBeat noted Anthropic couldn't enforce at the API layer, which is why global service collapsed. What broke the impasse was reportedly a personnel change: CNBC, citing a letter from Commerce Secretary Howard Lutnick, reports that co-founder Tom Brown took over the negotiations from CEO Dario Amodei. Mythos 5 was cleared for a set of approved U.S. organizations on June 26. Fable 5 followed on June 30.

The proximate trigger was an Amazon research report describing a safeguard bypass in which Fable 5 identified software vulnerabilities and, in one case, produced working exploit code. Anthropic's subsequent testing found the same behavior across Claude Haiku 4.5, Sonnet 4.6, Opus 4.6, 4.7, 4.8, GPT-5.4, GPT-5.5, and Kimi K2.7. The finding was industry-wide; the enforcement action wasn't.

Sonnet 5, meanwhile, is the more legible product story. TechCrunch reports it scores 63.2% on the agentic coding benchmark, against 69.2% for Opus 4.8 and 58.1% for Sonnet 4.6. Six points behind the flagship, at a fraction of the cost. Anthropic says it slightly outperforms Opus 4.8 on a knowledge-work benchmark, and calls its BrowseComp and OSWorld-Verified numbers a strict improvement over Sonnet 4.6 while spanning a wider cost-performance range than Opus 4.8.

"It can make plans, use tools like browsers and terminals, and run autonomously at a level that, just a few months ago, required larger and more expensive models," Anthropic wrote.

A few asterisks. Sonnet 5 uses an updated tokenizer, as Opus 4.7 did, that maps the same input to roughly 1.0–1.35× more tokens depending on content type. Anthropic says the introductory price is calibrated to make migration "roughly cost-neutral." Standard pricing of $3 input and $15 output kicks in after August 31. For Pro, Max, Team, and select Enterprise plans, Fable 5 gets a 50% included allowance of weekly usage limits through July 7, after which it moves to credits. AWS, Google Cloud, and Microsoft Foundry are re-enabling Fable 5 access "as quickly as possible."

The through-line worth naming: this is the first Claude frontier release that shipped through Commerce clearance before it shipped through Claude.ai. The 2019 Huawei entity-listing precedent applied that logic to hardware. It now applies to model weights.

Anthropic restores Claude Fable 5 globally, launches Sonnet 5 at $2/$10 introductory

Sources