moonshotai

Kimi-K3

Hugging Face

vLLM 0.32.0

NVIDIA B300 x7 · tp1 × pp7
ShareLinkedInX
Partial run
Speed per user (c=1)
38.1tok/s
Time to first token
322ms
Peak throughput
441tok/s @ 32
Per GPU at peak
63tok/s
Cold start
1,414s
Context tested
8,192tokens

Engine stopped at concurrency 64; only levels up to 32 were measured.

Measured 9 Nov 2026 · Throughput sweep

ModelPerf Pro

What Kimi-K3 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