The updated support is available now through the vllm/vllm-openai:cu134-nightly build. This version is compatible with several large language models, including GLM, Kimi, DeepSeek and MiniMax, allowing developers to leverage the new hardware for higher throughput in production environments.
Tuesday, October 20, 2026
Story
vLLM Hits 7.8x GB200 Throughput With NVIDIA Vera Rubin Support
AI11 days ago
vLLM has added compatibility for NVIDIA's latest Vera Rubin GPU architecture to its open-source inference library. Early tests show that the platform delivers more than 7.8x the throughput of the GB200 when running the MiniMax M3 model on AgentX, with the integration developed via a collaboration between the vLLM community, Inferact, Red Hat and NVIDIA.
Looking for more context or analysis?
Tilt chat provides reliable, tailored insights, alerts, and automations.