# vLLM update brings DeepSeek V4.1-Flash to eight A100 GPUs

> Source: <https://runtimewire.com/article/deepseek-v4-1-flash-now-runs-on-a100-in-vllm-at-489-p90-tok-s-per-user-on-h100>
> Published: 2026-11-17 18:41:30+00:00

# vLLM update brings DeepSeek V4.1-Flash to eight A100 GPUs

**Merged on November 16th, the decode path adds A100 support. The contributor reports 488.8 P90 tokens per second per user on H100 against an InferenceX line of 323.6, using different configurations.**

        By [RuntimeWire Staff](https://runtimewire.com/author/runtimewire-staff)
        · Published 
        · Updated 

DeepSeek-V4.1-Flash can now run on a single eight-A100 node in vLLM, after a specialized decode package from NVIDIA engineer Yaoyao Ding was merged into vLLM main on November 16th.

Ding's GitHub profile lists him as an engineer at NVIDIA who previously worked at CentML, AWS and MIT HAN Lab, with a computer science PhD from the University of Toronto and an ACM Honors Class background at Shanghai Jiao Tong University. The contribution is kernel work, not a product launch: before the merge, DeepSeek-V4.1-Flash did not run on A100.

Why A100 was locked out…
