Back to search results

Qwen3.5-27B · Q4_K_M

Start a discussion
NVIDIA RTX 509032GBllama.cpp b10741L1 Reproduced

This page aggregates 1 real-world runs of Qwen3.5-27B (Q4_K_M) on NVIDIA RTX 5090 with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

30.09 tok/s

Decode

Decode speed

1,515.83 tok/s

Prefill

Prefill speed

— s

TTFT

Time to first token

— GB

VRAM

VRAM usage

L1 Reproduced

1

Measured runs

1

Independent sources

GitHub

Source platforms

2026-10-02

Last verified

Performance

  1. Ollama · IQ3_XS Decode 206.43 · Prefill 827.13 ·
  2. Strata · Q2_K Decode 376.49 · Prefill 1620.15 ·
  3. Ollama · IQ1_S Decode 260.56 · Prefill 868.47 ·
  4. Ollama · IQ2_M Decode 235.26 · Prefill 1229.12 ·
  5. llama.cpp · IQ4_XS Decode 134 · Prefill 1478 ·
  6. llama.cpp · Q5_K_M Decode 114 · Prefill 1331 ·
  7. llama.cpp · Q8_0 Decode 92 · Prefill 792 ·
  8. vLLM · IQ4_XS Decode 108 · Prefill 1288 ·
  9. vLLM · Q4_K_M Decode 100 · Prefill 1182 ·
  10. vLLM · Q5_K_M Decode 106 · Prefill 956 ·
  11. vLLM · Q8_0 Decode 74 · Prefill 795 ·
  12. Ollama · IQ4_XS Decode 136 · Prefill 1332 ·
  13. Ollama · Q4_K_M Decode 125 · Prefill 1177 ·
  14. Ollama · Q5_K_M Decode 113 · Prefill 1178 ·
  15. Ollama · Q8_0 Decode 75 · Prefill 899 ·
  16. Strata · IQ4_XS Decode 198 · Prefill 1923 ·
  17. Strata · Q4_K_M Decode 178 · Prefill 1794 ·
  18. Strata · Q5_K_M Decode 147 · Prefill 1764 ·
  19. Strata · Q8_0 Decode 127 · Prefill 1290 ·
  20. llama.cpp · Q4_K_M (current)Decode 30.09 · Prefill 1515.83 ·

Core figures

Decode (avg)
30.09 tok/s
Prefill (avg)
1,515.83 tok/s
TTFT (avg)
— s
VRAM (avg)
— GB
MTP acceptance rate
—
TTFB
— GB
Power draw
— W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3.5-27B
Quantization
Q4_K_M
Framework
llama.cpp
Version
b10741

Hardware

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

GPU
NVIDIA RTX 5090
Nominal VRAM
32 GB
OS
Windows 11

Sources & evidence

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

  1. L1 ReproducedGitHubOriginal link Verified on 2026-10-02

    b10741 · Windows 11

    30.09 tok/s

    Decode

    1,515.83 tok/s

    Prefill

    — s

    TTFT

    0 GB

    VRAM

    —

    MTP

    — W

    Power draw

    qwen35 27B Q4_K_M | pp1024 | 1515.83 ± 47.03; qwen35 27B Q4_K_M | tg400 | 30.09 ± 0.44