DeepMind Scientist Tests Ox Alpha: Repeated Coding Failures

haider1 · x · 2026-08-22

A DeepMind research scientist discussed the new "Ox Alpha" model. Observations suggest that even at max effort, its coding performance is average: it realizes its mistake, apologizes and explains, but then repeats the exact same error. The behavior is reminiscent of old Gemini vibes.

Related event: DeepMind Researcher Hints Mystery Model Ox Alpha Is Next-Gen Gemini Pro(2 posts)→

Original post →

More from Models

Models channel →