OpenAI has launched GPT-6 Astra, a new flagship model billed as the first to reach the "Critical" cybersecurity capability tier under the company's own Preparedness Framework. The release arrived in the first week of September 2026 — just hours after Anthropic unveiled its own new flagship, Claude Fable 5.1, and its more constrained sibling Mythos 5.1.

In plain terms, OpenAI believes Astra can identify and respond to sophisticated cyber threats at a level no competitor has publicly demonstrated. Cybersecurity is one of the fastest-growing enterprise AI use cases, and claiming the highest internal safety tier for that domain could open doors with government and defense clients who have historically been cautious about adopting frontier models.

The Astra launch is the capstone of an aggressive product push that began in July 2026, when OpenAI rolled out its GPT-5.6 family — a trio of models branded Sol, Terra and Luna, aimed at different performance and cost tiers, with OpenAI claiming state-of-the-art results across several coding benchmarks.

The stakes are high. By mid-2026, Anthropic had overtaken OpenAI in annualized revenue, a shift driven largely by Claude Code, Anthropic's developer-focused product that became a runaway hit with engineering teams. OpenAI CEO Sam Altman has publicly acknowledged past strategic missteps that left the company trailing during the 2025–2026 stretch — and the Astra and GPT-5.6 launches represent its most aggressive attempt yet to reclaim the lead.

Anthropic responded to the pressure on pricing: its September releases came with a roughly 25% price cut on standard workloads for Claude Fable 5.1, aimed at enterprise customers who weigh cost-per-token as heavily as raw capability.

One caveat matters. OpenAI's claim of having overtaken Anthropic rests on its own benchmarks and its own safety framework — both self-reported. Anthropic's revenue lead, meanwhile, is a lagging indicator that may not yet reflect the impact of OpenAI's newest launches.

Sources