schema: onefrontier.launch-card/v1
hf:
  repo: Qwen/Qwen3.8-27B
  revision: 1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0
  license: apache-2.0
  gated: false
  params_b: 27.78
  modality:
  - image-text-to-text
  weight_gib: 51.7
  last_modified: '2026-08-14T15:00:01.000Z'
  downloads: 7581050
checkpoints:
- name: root
  format: bf16
  subdir: null
  source_repo: null
  resident_gib: 51.7
  host_gib: null
  gated: false
  notes: 18 weight file(s)
envs:
- engine: vllm
  checkpoint: root
  engine_version: null
  image: vllm/vllm-openai:qwen38
  argv:
  - vllm
  - serve
  - Qwen/Qwen3.8-27B
  - --tensor-parallel-size
  - '1'
  - --enable-auto-tool-choice
  - --tool-call-parser
  - qwen3_coder
  - --reasoning-parser
  - qwen3
  - --mm-encoder-tp-mode
  - data
  command: "vllm serve Qwen/Qwen3.8-27B \\\n  --tensor-parallel-size 1 \\\n  --enable-auto-tool-choice\
    \ \\\n  --tool-call-parser qwen3_coder \\\n  --reasoning-parser qwen3 \\\n  --mm-encoder-tp-mode\
    \ data"
  env:
    VLLM_USE_RUST_FRONTEND: '1'
  validated_hardware:
  - gb300
  - b200
  - rtx_5090
  - rtx_5090_2x
  - rtx_pro_6000
  - dgx_spark_gb10
  - ascend_950pr
  - arc_pro_b60
  - arc_pro_b70
  notes: null
  source:
    kind: vllm-recipes
    url: https://recipes.vllm.ai/Qwen/Qwen3.8-27B.json
    fetched_at: null
hardware:
  gpu_models:
  - b200
  - gb300
  - rtx_5090
  - rtx_5090_2x
  - rtx_pro_6000
  gpu_arch:
  - blackwell
  min_gpus: 1
  tp_rule: widest published TP is 1
  min_vram_per_gpu_gib: null
  min_total_vram_gib: null
  min_disk_gib: 112
  min_host_ram_gib: null
  interconnect: null
  cuda_min: null
  driver_min: null
  context_tokens: null
performance: https://onefrontier.ai/Qwen/Qwen3.8-27B/performance
evidence:
- kind: hf-api
  url: https://huggingface.co/api/models/Qwen/Qwen3.8-27B
  fetched_at: null
- kind: vllm-recipes
  url: https://recipes.vllm.ai/Qwen/Qwen3.8-27B.json
  fetched_at: null
- kind: sglang-cookbook
  url: https://docs.sglang.io/cookbook/autoregressive/Qwen/Qwen3.8-27B.md
  fetched_at: null
gaps:
- field: envs.sglang
  why: 'SGLang cookbook page exists but its launch command is generated by the page''s
    Deploy widget, not printed in the markdown; the page documents these flags: --mamba-full-memory-ratio,
    --speculative-num-draft-tokens, --speculative-dspark-block-size, --enable-linear-replayssm-spec,
    --mamba-ssm-dtype, --max-mamba-cache-size, --max-running-requests, --mem-fraction-static,
    --max-total-tokens, --kv-cache-dtype, --attention-backend, --device, --speculative-algorithm,
    --tp-size...'
- field: hardware.npu
  why: 'recipes name non-GPU accelerators too: arc_pro_b60, arc_pro_b70, ascend_950pr,
    dgx_spark_gb10 (not carded)'
captured_at: '2026-10-31T17:28:18Z'
captured_by: hf-api + vllm-recipes + sglang-cookbook