Anthropic has marked a pivotal week with three major developments: the release of a powerful new agentic model, the launch of a dedicated scientific research workbench, and the restoration of global access to its most advanced AI systems following the lifting of U.S. export restrictions.
The U.S. government has removed its requirement that Anthropic obtain export licenses for its Mythos and Fablemodels, with access restored from Wednesday, July 1. The models had been restricted since June 12, when they were added to a list of export-controlled technologies, effectively ending public availability. Secretary of Commerce Howard Lutnick confirmed the reversal, stating that Anthropic had agreed to proactively detect and address security risks, cooperate with the U.S. government on protocols for future model releases, and report any malicious activity. Pressure to lift the ban had grown as Asian competitors including Fugu and Tulongfeng began releasing models approaching Mythos-level capability.
Separately, Anthropic launched Claude Sonnet 5, positioning it as its most capable mid-tier model to date and the new default for free and Pro subscribers. The model delivers agentic performance, including autonomous reasoning, tool use, and coding, at a cost significantly below Opus 4.8, priced at $2 per million input tokens through August. Daniel Shepardof Zapier and Fabian Hedin of Lovable both described the model as reliably completing complex multi-step tasks that previously stalled.
Anthropic also unveiled Claude Science, an AI research workbench connecting scientists to over 60 databases with built-in fact-checking and reproducibility tools. Jérôme Lecoq of the Allen Institute and researchers at the UCSF Brain Tumor Center are already using it in production workflows.
