OpenAI Launches GPT‑6 Astra Ultrafast Powered by NVIDIA Blackwell GPUs

OpenAI has rolled out GPT‑6 Astra Ultrafast, a new variant of its flagship language model that is now available in the OpenAI API and for eligible users of ChatGPT Work and Codex.

The upgrade is built on NVIDIA’s Blackwell GPUs. By tuning inference workloads to the architecture’s strengths, OpenAI reports up to eight times faster token generation compared to the standard Astra mode.

Faster inference means lower latency for real‑time applications and a potential reduction in compute costs for developers. The speed boost could help businesses deploy more responsive chatbots, code assistants, and other AI‑driven services.

Developers can access Astra Ultrafast through the same API endpoints used for earlier models, while eligible ChatGPT Work and Codex users can try the new speed tier directly in their existing workflows. The move underscores OpenAI’s focus on performance as a key differentiator for its AI services.