All issues

Search results

MiMo-V2.6-Flash-RL also runs on 4x AMD MI210 (gfx90a / CDNA2) — measured notes for the recipe

Status: Open.

#1046 In vllm-project/recipes;

· gajanandjha opened on Sep 29, 2026

Kimi-K3 B200 pd_cluster: DEP16/TP1 decode appears to exceed 180 GB per GPU

Status: Open.

#1030 In vllm-project/recipes;

· ofer1katz opened on Sep 24, 2026

GLM-5.3 AMD command fails on vllm-openai-rocm:latest (v0.30.0) until #57252 ships in a release

Status: Open.

#1027 In vllm-project/recipes;

· JPedro2 opened on Sep 24, 2026

[DeepSeek-V4-Flash-0731] DSpark on 4× RTX PRO 6000: num_speculative_tokens: 3 measured ahead of the default 7 under a multi-agent load

Status: Open.

#967 In vllm-project/recipes;

· J-Oe-2 opened on Sep 13, 2026

docker save exports only 18K tar for deepseekv41-flash-0909 (no layer blobs)

Status: Open.

#955 In vllm-project/recipes;

· musica2016 opened on Sep 11, 2026

[Usage] Qwen3.8-27B support on DGX Spark (GB10)

Status: Open.

#939 In vllm-project/recipes;

· xiabo0816 opened on Sep 8, 2026

Fixed request seed produces 5x more exact action cycles with Qwen3.8 27B FP8

Status: Open.

#925 In vllm-project/recipes;

· vvsotnikov opened on Sep 3, 2026

microsoft/Phi-4-reasoning Xeon 6 recipe incomplete

Status: Open.

#912 In vllm-project/recipes;

· jharriga opened on Sep 1, 2026

microsoft/Phi-4-multimodal-instruct Xeon 6 recipe incomplete

Status: Open.

#911 In vllm-project/recipes;

· jharriga opened on Sep 1, 2026

zai-org/glm-4-9b-hf Xeon 6 recipe incomplete

Status: Open.

#909 In vllm-project/recipes;

· jharriga opened on Sep 1, 2026

openai/whisper-large-v3 Xeon 6 recipe incomplete

Status: Open.

#908 In vllm-project/recipes;

· jharriga opened on Sep 1, 2026

[Bug] multi_node_tp_pp + GLM-5.2 + MTP: synthesized command fails at startup with NotImplementedError

Status: Open.

#875 In vllm-project/recipes;

· aminer-404 opened on Aug 28, 2026