Anthropic Ships Claude Sonnet 5, Hours Before Washington Lifts Its Mythos Export Ban
A single week brought Anthropic both a faster, cheaper AI model and the return of its most powerful ones, as the Commerce Department dropped restrictions that had cut off Claude Mythos 5 and Fable 5 from the world.
Anthropic had an unusually eventful Tuesday. The company introduced Claude Sonnet 5, a mid-tier model built for agentic work like coding, browsing, and multi-step task execution, pricing it well below its flagship Opus 4.8.
Hours later, the Commerce Department notified Anthropic that it was removing export controls imposed against Claude Mythos 5 and Fable 5, restrictions that had forced the company to pull both models from public access entirely.
Together, the two developments capped a turbulent stretch for the San Francisco AI lab, one defined by rapid product releases and an unusually tense back-and-forth with federal regulators.
Sonnet 5 Targets the Agentic Middle Ground
Claude Sonnet 5 is Anthropic’s most agentic Sonnet model yet, capable of making plans, operating tools such as browsers and terminals, and running autonomously at a level that previously required larger, costlier models.
Anthropic says its performance approaches Opus 4.8 at a lower price, and TechCrunch reports the model is now the default for Free and Pro Claude.ai users while remaining available across Max, Team, and Enterprise tiers, Claude Code, and the Claude Platform.
Sonnet 5 launches at $2 per million input tokens and $10 per million output tokens through August 31, after which pricing rises to $3 and $15, respectively, undercutting both Opus 4.8 and higher-priced offerings from OpenAI and Google.
Benchmarks cited by TechCrunch showed Sonnet 5 scoring 63.2% on an agentic coding test, trailing Opus 4.8’s 69.2% but well ahead of Sonnet 4.6’s 58.1%, while on a separate knowledge-work benchmark it edged past Opus 4.8 itself.
Zapier senior engineer Daniel Shepard described handing the model a two-part Salesforce and outreach task that it completed end to end, calling it “a no-brainer” for daily automation.
While Fabian Hedin, the co-founder of Swedish tech firm Lovable, praised its consistency in refusing unsafe requests.
Washington Reverses an 18-Day Blackout
The export controls dated back to June 12, when the Commerce Department required Anthropic to obtain government approval before allowing any foreign national access to Mythos 5 or Fable 5.
The company found this mandate impossible to enforce at scale, and that led it to shut both models down entirely. The decision drew criticism from cybersecurity researchers, who argued it was less a genuine safety measure than leverage against Anthropic.
The company had publicly disputed with the administration, including through a lawsuit against the Pentagon earlier that year over surveillance and autonomous weapons assurances.
Commerce Secretary Howard Lutnick said his department worked closely with Anthropic to “analyze and approve” the models.
He credited the company with agreeing to detect and address security risks proactively, coordinate future release protocols, and report malicious activity.
Anthropic confirmed access would begin returning today, July 1, framing the resolution as the product of weeks of technical and policy discussions with federal officials.
Competitive Pressure Shaped the Timing
Rival labs OpenAI and Google have pushed their own agentic systems, GPT-5.6 Sol and Gemini 3.5 Flash, into the market in recent weeks, reinforcing the industry’s shift toward autonomous, tool-using models as a baseline rather than a differentiator.
Meanwhile, Asian AI developers behind models such as Tulongfeng have narrowed the capability gap with Mythos-class systems, adding pressure on U.S. regulators to avoid sidelining American labs as global competition in frontier AI intensifies.
Anthropic had already cleared Mythos 5 for a limited set of government-approved organizations the week before the broader reversal, suggesting officials were easing restrictions in stages rather than all at once.
With the export ban lifted and Sonnet 5 now generally available, Anthropic enters July with its full model lineup restored and a new agentic contender aimed squarely at cost-conscious developers.
Source: Introducing Claude Sonnet 5
