Qwen3.8-Flash-Next
serving recipe.

Floors are the minimum the vLLM recipe publishes across the whole replica, from vLLM's own runs. The control plane compares them with its own sizing from the model config.

• Tensor ParallelFits inside one node's interconnect. single_node_tp
• Tensor + Expert ParallelFits inside one node's interconnect. single_node_tep
• Data + Expert ParallelFits inside one node's interconnect. single_node_dep

• DGX Station (GB300) · 1 GPU
• GB200 NVL4 · 4 GPUs
• GB300 NVL4 · 4 GPUs
• H200 · 8 GPUs
• RTX Pro 6000 4x · 4 GPUs
• MI355X · 8 GPUs

Wingsmith runs models through NVIDIA Dynamo, so it serves on NVIDIA GPUs. Other vendors are listed as the recipe states them.

vllm serve Inferact/Qwen3.8-Flash-Next-NVFP4 \
  --max-num-seqs 256 \
  --gpu-memory-utilization 0.90 \
  --enable-prefix-caching \
  --no-enable-flashinfer-autotune \
  --enable-auto-tool-choice \
  --tool-call-parser qwen3_xml \
  --reasoning-parser qwen3

Features

• Tool callingOn by defaultEnable automatic tool choice with the Qwen3 XML parser
• ReasoningOn by defaultExtract thinking output with the Qwen3 reasoning parser
• Speculative decodingOpt-inUse the checkpoint's Multi-Token Prediction module for faster decoding
• Text-only modeOpt-inSkip the vision encoder for text-only workloads and leave more memory for KV cache

1. Register the weights

   In the console, or through the API. Resolving a slug proposes an artifact and writes nothing.

   curl "http://localhost:8000/v1/artifacts/resolve?source=hf:Qwen/Qwen3.8-Flash-Next"

2. Read the draft

   The draft carries this recipe as upstream evidence: the tested flags, the precision and memory floor, the parallelism layouts, and the hardware vLLM verified, with links back to the recipe. Nothing from it is applied without you, and the draft warns when the plane's own sizing and the published floor differ by a quarter or more.

3. Create the workload

   Choose the engine, device shape and scale. The plane picks the clusters, and each cluster's operator renders the deployment as a Dynamo graph.

4. Give it an endpoint

   One model name for clients, routed by the gateway to deployments on every cluster serving it.

Run the control plane locally →

• How much GPU memory does Qwen3.8-Flash-Next need?The vLLM recipe publishes a minimum of 130 GB across the whole replica for its NVFP4 variant; FP8 250 GB, BF16 423 GB. That is vLLM's own measurement, not a per-GPU figure.
• Which parallelism layouts does the Qwen3.8-Flash-Next recipe support?Tensor Parallel, Tensor + Expert Parallel and Data + Expert Parallel.
• Can Wingsmith serve Qwen3.8-Flash-Next?Wingsmith serves models through NVIDIA Dynamo on clusters with NVIDIA GPUs. Register Qwen/Qwen3.8-Flash-Next and the plane attaches this vLLM recipe to the draft as evidence: its tested flags, precision and memory floor, for you to review rather than applied on your behalf.

From the vLLM recipe for Qwen/Qwen3.8-Flash-Next ↗ (snapshot at 4bcc189), Apache 2.0.