Arao AI
Back to search results

Qwen3.5-27B · Q4_K_M

Start a discussion
Intel Arc A770 16GB16GBllama.cpp b4213L2 Cross-framework verified

This page aggregates 1 real-world runs of Qwen3.5-27B (Q4_K_M) on Intel Arc A770 16GB with llama.cpp, contributed by 1 independent source platforms; metrics are averages of published measurements.

77 tok/s

Decode

Decode speed

681 tok/s

Prefill

Prefill speed

1.78 s

TTFT

Time to first token

15.5 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 84 · Prefill 777 ·
  2. llama.cpp · Q4_K_M (current)Decode 77 · Prefill 681 ·
  3. llama.cpp · Q5_K_M Decode 68 · Prefill 739 ·
  4. llama.cpp · Q8_0 Decode 47 · Prefill 559 ·
  5. vLLM · IQ4_XS Decode 69 · Prefill 706 ·
  6. vLLM · Q4_K_M Decode 62 · Prefill 655 ·
  7. vLLM · Q5_K_M Decode 54 · Prefill 646 ·
  8. vLLM · Q8_0 Decode 45 · Prefill 434 ·
  9. Ollama · IQ4_XS Decode 66 · Prefill 791 ·
  10. Ollama · Q4_K_M Decode 62 · Prefill 744 ·
  11. Ollama · Q5_K_M Decode 68 · Prefill 680 ·
  12. Ollama · Q8_0 Decode 50 · Prefill 494 ·
  13. Strata · IQ4_XS Decode 115 · Prefill 1228 ·
  14. Strata · Q4_K_M Decode 105 · Prefill 1087 ·
  15. Strata · Q5_K_M Decode 100 · Prefill 950 ·
  16. Strata · Q8_0 Decode 63 · Prefill 752 ·

Core figures

Decode (avg)
77 tok/s
Prefill (avg)
681 tok/s
TTFT (avg)
1.78 s
VRAM (avg)
15.5 GB
MTP acceptance rate
—
TTFB
— GB
Power draw
326 W

Configuration

Member-level fields are taken from the most recent run

Model
Qwen3.5-27B
Quantization
Q4_K_M
Framework
llama.cpp
Version
b4213
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 A770 16GB
Nominal VRAM
16 GB
Measured VRAM (avg)
15.5 GB
OS
Linux 6.8.0
Driver
570.86.15
CUDA
12.6
Power draw
326 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

    b4213 · Linux 6.8.0 · CUDA 12.6 · 8192 ctx

    77 tok/s

    Decode

    681 tok/s

    Prefill

    1.78 s

    TTFT

    15.5 GB

    VRAM

    —

    MTP

    326 W

    Power draw

    Auto-generated LLM-sourced record 2026-10-09. Model: Qwen3.5-27B, Hardware: Arc A770, Framework: 1, Quant: Q4_K_M.