The GPT-5 generation consolidated multimodal reasoning and agents, raising the baseline users and enterprises expect.
By late 2025, OpenAI and rivals launched GPT-5-class systems with better tool use, product memory, and reasoning. Training details stayed opaque.
Saturated benchmarks pushed new long-horizon and real-economy evaluations. Focus moved to reliability on 30–100 step flows.
That generation prepared the ground for Astra and more capable internal systems announced in 2026.
OpenAI presented GPT-6 Astra as a more capable intelligence generation; later internal systems would be cited in the Navier–Stokes announcement.
OpenAI's new multimodal models integrate text, image, and audio with refreshed APIs for developers.
OpenAI describes ~10,000 concurrent agents, hundreds of billions of tokens, and Lean formalization after finding a singularity with smooth forcing.