Back to search results

Qwen3-8B · Q8_0

Start a discussion
NVIDIA RTX 409024GBllama.cpp v0.29L2 Cross-framework verified

This page aggregates 3 real-world runs of Qwen3-8B (Q8_0) on NVIDIA RTX 4090 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

87.18 tok/s

Decode

Decode speed

465.1 tok/s

Prefill

Prefill speed

1.06 s

TTFT

Time to first token

17.31 GB

VRAM

VRAM usage

L2 Cross-framework verified

3

Measured runs

1

Independent sources

Other

Source platforms

2026-10-18

Last verified

Performance

  1. llama.cpp · Q4_K_M Decode 142.71 · Prefill 813.73 ·
  2. vLLM · IQ3_S Decode 161.14 · Prefill 813.02 ·
  3. llama.cpp · IQ1_S Decode 187.88 · Prefill 983.15 ·
  4. llama.cpp · Q8_0 (current)Decode 87.18 · Prefill 465.1 ·
  5. vLLM · Q2_K Decode 164.91 · Prefill 671.93 ·
  6. llama.cpp · Q6_K Decode 88.33 · Prefill 428.17 ·
  7. vLLM · Q4_K_S Decode 126.59 · Prefill 467.91 ·
  8. vLLM · Q8_0 Decode 76.65 · Prefill 513.21 ·
  9. Ollama · Q6_K Decode 82.66 · Prefill 332.47 ·
  10. llama.cpp · IQ4_XS Decode 112 · Prefill 1323 ·
  11. vLLM · IQ4_XS Decode 104 · Prefill 1176 ·
  12. vLLM · Q4_K_M Decode 96 · Prefill 1061 ·
  13. vLLM · Q5_K_M Decode 95 · Prefill 815 ·
  14. Ollama · Q4_K_M Decode 113 · Prefill 922 ·
  15. Ollama · Q5_K_M Decode 95 · Prefill 1090 ·
  16. llama.cpp · Q5_K_M Decode 133.3 · Prefill 999.8 ·

Core figures

Decode (avg)
87.18 tok/s
Prefill (avg)
465.1 tok/s
TTFT (avg)
1.06 s
VRAM (avg)
17.31 GB
MTP acceptance rate
—
TTFB
0.68 GB
Power draw
386.4 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3-8B
Quantization
Q8_0
Framework
llama.cpp
Version
v0.29
Context length
16384 tokens
Batch size
512
GPU layers
28
Flash Attention
On

Hardware

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

GPU
NVIDIA RTX 4090
Nominal VRAM
24 GB
Measured VRAM (avg)
17.31 GB
OS
Ubuntu 24.04
Driver
560.94
CUDA
12.8
Power draw
301.7 W

Sources & evidence

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

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

    v0.29 · Ubuntu 24.04 · CUDA 12.8 · 16384 ctx

    82.94 tok/s

    Decode

    405.82 tok/s

    Prefill

    1.07 s

    TTFT

    21.42 GB

    VRAM

    —

    MTP

    301.7 W

    Power draw

    L2 batch seed 2026-10-18: llama.cpp on hw_id=1 running model_id=53 at Q8_0
  2. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-10

    v0.28 · Ubuntu 24.04 · CUDA 12.8 · 2048 ctx

    87.21 tok/s

    Decode

    303.99 tok/s

    Prefill

    1.97 s

    TTFT

    21.42 GB

    VRAM

    —

    MTP

    471.1 W

    Power draw

    L2 batch seed 2026-10-10: llama.cpp on hw_id=1 running model_id=53 at Q8_0
  3. L0 Self-reportedOtherOriginal link Verified on 2026-10-06

    b4200 · Linux · CUDA 12.6 · 4096 ctx

    91.4 tok/s

    Decode

    685.5 tok/s

    Prefill

    0.15 s

    TTFT

    9.1 GB

    VRAM

    —

    MTP

    — W

    Power draw

    L0 理论估算(add-1000-records-batch3)