Arao AI
Back to search results

Kimi-K3 · Q3_K_L

Start a discussion
NVIDIA RTX 507012GBvLLM v0.53L2 Cross-framework verified

This page aggregates 1 real-world runs of Kimi-K3 (Q3_K_L) on NVIDIA RTX 5070 with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.

76.49 tok/s

Decode

Decode speed

291.62 tok/s

Prefill

Prefill speed

0.76 s

TTFT

Time to first token

1.05 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

Today

Last verified

Performance

  1. vLLM · Q3_K_L (current)Decode 76.49 · Prefill 291.62 ·
  2. Strata · Q5_K_M Decode 76.48 · Prefill 397.87 ·
  3. SGLang · Q5_K_M Decode — · Prefill — ·
  4. SGLang · Q4_K_M Decode — · Prefill — ·
  5. vLLM · Q5_K_M Decode — · Prefill — ·
  6. vLLM · Q4_K_M Decode — · Prefill — ·

Core figures

Decode (avg)
76.49 tok/s
Prefill (avg)
291.62 tok/s
TTFT (avg)
0.76 s
VRAM (avg)
1.05 GB
MTP acceptance rate
—
TTFB
0.61 GB
Power draw
231.7 W

Configuration

Member-level fields are taken from the most recent run

Model
Kimi-K3
Quantization
Q3_K_L
Framework
vLLM
Version
v0.53
Context length
32768 tokens
Batch size
8
GPU layers
999
Flash Attention
Off

Hardware

Nominal and measured figures are shown side by side; whether it runs is the reader's call

GPU
NVIDIA RTX 5070
Nominal VRAM
12 GB
Measured VRAM (avg)
1.05 GB
OS
Ubuntu 24.04
Driver
560.94
CUDA
12.4
Power draw
231.7 W

Sources & evidence

1 measured records in total, each traceable to its original source

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-18

    v0.53 · Ubuntu 24.04 · CUDA 12.4 · 32768 ctx

    76.49 tok/s

    Decode

    291.62 tok/s

    Prefill

    0.76 s

    TTFT

    1.05 GB

    VRAM

    —

    MTP

    231.7 W

    Power draw

    L2 batch seed 2026-10-18: vLLM on hw_id=18 running model_id=73 at Q3_K_L