Qwen

Qwen3.8-27B

Hugging Face

SGLang 0.5.22

NVIDIA B300 x1 · tp1 × pp1
ShareLinkedInX
Speed per user (c=1)
95.6tok/s
Time to first token
63ms
Peak throughput
4,052tok/s @ 128
Per GPU at peak
4,052tok/s
Cold start
158s
Context tested
8,192tokens

Measured 9 Nov 2026 · Throughput sweep

vLLM 0.32.0

NVIDIA B300 x1 · tp1 × pp1
ShareLinkedInX
Eco-friendly
Speed per user (c=1)
91.6tok/s
Time to first token
69ms
Peak throughput
5,947tok/s @ 256
Per GPU at peak
5,947tok/s
Cold start
85s
Context tested
8,192tokens

Measured 9 Nov 2026 · Throughput sweep

ModelPerf Pro

What Qwen3.8-27B costs to run, per token and per watt

  • Cost per million tokens for every model, engine and GPU
  • Energy per million tokens, tokens per watt and GPU power draw
  • Full concurrency curves for all three workload scenarios
  • Unlimited ROI and break-even calculations with PDF export
  • CSV export of the full leaderboard, paid metrics included
  • New models and re-runs included as they are published
$39 / month, billed annually

or $49 month to month

Get ModelPerf ProSign in to buy · $10Compare plans