Arao AI
Back to search results

Qwen3.5-27B · Q8_0

Start a discussion
Intel Arc B58012GBvLLM 0.7.3L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3.5-27B (Q8_0) on Intel Arc B580 with vLLM, contributed by 1 independent source platforms; metrics are averages of published measurements.

31 tok/s

Decode

Decode speed

346 tok/s

Prefill

Prefill speed

0.98 s

TTFT

Time to first token

31.1 GB

VRAM

VRAM usage

L2 Cross-framework verified

1

Measured runs

1

Independent sources

Other

Source platforms

Today

Last verified

Performance

  1. llama.cpp · IQ4_XS Decode 58 · Prefill 632 ·
  2. llama.cpp · Q4_K_M Decode 53 · Prefill 561 ·
  3. llama.cpp · Q5_K_M Decode 51 · Prefill 470 ·
  4. llama.cpp · Q8_0 Decode 32 · Prefill 380 ·
  5. vLLM · IQ4_XS Decode 51 · Prefill 428 ·
  6. vLLM · Q4_K_M Decode 46 · Prefill 405 ·
  7. vLLM · Q5_K_M Decode 38 · Prefill 447 ·
  8. vLLM · Q8_0 (current)Decode 31 · Prefill 346 ·
  9. Ollama · IQ4_XS Decode 52 · Prefill 592 ·
  10. Ollama · Q4_K_M Decode 46 · Prefill 533 ·
  11. Ollama · Q5_K_M Decode 46 · Prefill 524 ·
  12. Ollama · Q8_0 Decode 37 · Prefill 297 ·
  13. Strata · IQ4_XS Decode 76 · Prefill 894 ·
  14. Strata · Q4_K_M Decode 70 · Prefill 812 ·
  15. Strata · Q5_K_M Decode 73 · Prefill 616 ·
  16. Strata · Q8_0 Decode 49 · Prefill 546 ·

Core figures

Decode (avg)
31 tok/s
Prefill (avg)
346 tok/s
TTFT (avg)
0.98 s
VRAM (avg)
31.1 GB
MTP acceptance rate
—
TTFB
— GB
Power draw
472 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3.5-27B
Quantization
Q8_0
Framework
vLLM
Version
0.7.3
Context length
8192 tokens
Batch size
512
GPU layers
99
Flash Attention
On

Hardware

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

GPU
Intel Arc B580
Nominal VRAM
12 GB
Measured VRAM (avg)
31.1 GB
OS
Linux 6.8.0
Driver
570.86.15
CUDA
12.6
Power draw
472 W

Sources & evidence

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

  1. L2 Cross-framework verifiedOtherOriginal link Verified on 2026-10-09

    0.7.3 · Linux 6.8.0 · CUDA 12.6 · 8192 ctx

    31 tok/s

    Decode

    346 tok/s

    Prefill

    0.98 s

    TTFT

    31.1 GB

    VRAM

    —

    MTP

    472 W

    Power draw

    Auto-generated LLM-sourced record 2026-10-09. Model: Qwen3.5-27B, Hardware: Arc B580, Framework: 2, Quant: Q8_0.