OpenAI Seeks to Reclaim AI Leadership with Cybersecurity-Focused Astra

OpenAI has claimed to have overtaken Anthropic in the AI race, thanks largely to its newest model Astra and the benchmarking results of its GPT-5.6 family.

Astra was launched on September 2, 2026, as a model that can reach the 'Critical' cybersecurity capability tier under OpenAI's internal Preparedness Framework.

This means OpenAI believes Astra can identify and respond to sophisticated cyber threats at a level no competitor has publicly demonstrated.

Anthropic had unveiled Claude Fable 5.1 and Mythos 5.1 just hours earlier, with a notable sweetener of a 25% price cut for standard workloads on Claude Fable 5.1.

OpenAI's GPT-5.6 family was rolled out two months earlier in July 2026, with three variants aimed at different performance and cost tiers, claiming state-of-the-art results across several coding benchmarks.