{
  "id": 11311456,
  "title": "How NVIDIA GPUs Help Accelerate OpenAI’s GPT-6 Astra Ultrafast",
  "url": "https://urgent.news/2026/10/01/how-nvidia-gpus-help-accelerate-openais-gpt-6-astra-ultrafast",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-01T23:44:13.000Z",
  "source": {
    "name": "NVIDIA Blog",
    "slug": "nvidia-blog",
    "url": "https://blogs.nvidia.com/blog/gpus-openai-gpt-6-astra-ultrafast/"
  },
  "original_language": "en",
  "account": "GPT-6 Astra Ultrafast, powered by NVIDIA Blackwell GPUs, is now accessible via the OpenAI API and to certain ChatGPT Work and Codex users. This accelerated version delivers up to eight times faster token generation compared to the Astra Standard mode. For developers, this translates to shorter edit-test-debug cycles, reduced response time between tool calls, and enhanced interactivity in applications. The impact of Astra Ultrafast becomes particularly evident in time-sensitive loops, such as when an agent writes code, utilizes a tool, evaluates the outcome, and determines the next action.\n\nNVIDIA AI infrastructure plays a crucial role in enabling OpenAI to deliver more useful model outputs when developers require them. Philippe Tillet, inference lead at OpenAI, highlighted NVIDIA's significant investment in tooling and documentation, which has allowed OpenAI to create models proficient in programming Blackwell and Rubin GPUs. Astra can leverage this expertise to develop high-performance kernels that ensure NVIDIA hardware's competitiveness across various latency, throughput, and cost scenarios. With Astra Ultrafast, this means faster model responses as agents engage in coding, tool usage, and completing complex tasks.\n\nOpenAI's commitment to continuous performance enhancement is evident in its utilization of its own models to refine the inference software running on NVIDIA GPUs. By taking advantage of the platform's programmability, OpenAI can test and implement improvements, leading to faster model responses and a more productive infrastructure over time. Uday Ruddarraju, chief technology officer of compute at OpenAI, emphasized that their collaboration with NVIDIA enables them to make AI faster and more useful. The internal models were employed to optimize inference on NVIDIA GPUs, while NVIDIA's programmability facilitated the delivery of acceleration behind Astra Ultrafast.\n\nThe programmable NVIDIA platform offers developers and researchers the flexibility to reuse infrastructure across different stages of AI development, including training, inference, and reinforcement learning. This adaptability allows teams to repurpose compute resources as demand fluctuates, improving resource utilization and preventing unnecessary overprovisioning for specific workloads. Access to GPT-6 Astra Ultrafast is currently available through the OpenAI API. Developers can find more information regarding access, pricing, and implementation details in the Ultrafast guide.",
  "summary": "GPT-6 Astra Ultrafast, running on NVIDIA Blackwell GPUs, is available now in the OpenAI API and to eligible ChatGPT Work and Codex users. Accelerated by inference optimizations through OpenAI’s models that tap into the capabilities of the NVIDIA Blackwell architecture, Ultrafast offers up to 8x faster token generation than the Astra Standard mode. For developers, […]",
  "key_points": [
    "GPT-6 Astra Ultrafast powered by NVIDIA Blackwell GPUs",
    "Up to eight times faster token generation than Astra Standard",
    "OpenAI API and ChatGPT Work and Codex users can access Ultrafast"
  ],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}