# Recipe: Qwen3-Coder-Next-FP8
# Qwen3-Coder-Next model in native FP8 format
# Currently can only be run in solo mode, cluster mode fails with error - tracking https://github.com/vllm-project/vllm/issues/33857

recipe_version: "1"
name: Qwen3-Coder-Next-FP8
description: vLLM serving Qwen3-Coder-Next-FP8 on a SINGLE NODE ONLY!

# HuggingFace model to download (optional, for --download-model)
model: Qwen/Qwen3-Coder-Next-FP8

# This model can only run on single node (solo)
solo_only: true

# Container image to use
container: vllm-node

# No mods required
mods: []

# Default settings (can be overridden via CLI)
defaults:
  port: 8000
  host: 0.0.0.0
  tensor_parallel: 1
  gpu_memory_utilization: 0.7
  max_model_len: 131072

# Environment variables
env: {}

# The vLLM serve command template
command: |
  vllm serve Qwen/Qwen3-Coder-Next-FP8 \
    --enable-auto-tool-choice \
    --tool-call-parser qwen3_coder \
    --gpu-memory-utilization {gpu_memory_utilization} \
    --host {host} \
    --port {port} \
    --load-format fastsafetensors \
    --attention-backend flashinfer \
    --enable-prefix-caching \
    --max-model-len {max_model_len}