deepseek-v4.1POST
vLLM + 蚂蚁 Humming：8x H20 上 V4.1 Flash 吞吐 +14%/+43%
vLLM 与蚂蚁开源 JIT 量化 GEMM 内核 Humming，在 8x H20 上跑 V4.1 Flash（MXFP4 权重、FP8 激活），并发 1 吞吐 +14%，并发 64 时 +43%。
海猿
@guolinwang10 天前收录原推发布于 2026.10.22
海猿 @guolinwang原推发布于 2026.10.22
vLLM + Ant Group open-sourced Humming, JIT quantized GEMM kernels. DeepSeek-V4.1-Flash on 8x H20, same MXFP4 weights, FP8 activations: +14% throughput at concurrency 1, +43% at 64. Big batches are compute-bound, so 4-bit weights alone stop paying. https://vllm.ai/blog/2026-10-21-humming
x.com/guolinwang/status/2113030967929520346采集于 2026.10.22 12:01