Astra and Opus just passed Turing’s other test
Claude Opus and Google Astra just cracked what Turing left unfinished—and it says something about how far reasoning models have come.
Why it matters
Two frontier models have demonstrated capability on a historical codebreaking benchmark framed as an extension of Turing's work. This is a substantive capability milestone for reasoning-class models, though the article's framing requires scrutiny: the claim hinges on what 'passing' means, whether the benchmark is independently validated, and how it compares to human performance or prior baselines.
The key facts
5 to knowAstra and Opus (Claude 3.5 Opus variant) both passed the benchmark
Benchmark framed as completing Alan Turing's World War II codebreaking work
Published Sep 25, 2026 on TechCrunch
No pricing, region, or consumption-unit details provided in headline
No independent validation source cited in headline; benchmark design and success criteria unclear
Go to the source
TechCrunch AItechcrunch.com
Publisher excerpt: Frontier AI models are finishing Alan Turing's World War II codebreaking work.