[Bug]: Qwen3-Coder-30B-A3B-Instruct-int4-ov runs on CPU but triggers OOM on iGPU

[ikirsh]

OpenVINO Version

OpenVINO Model Server 2026.0.0.4d3933c5c, OpenVINO backend 2026.0.0.0rc3

Operating System

Other (Please specify in description)

Device used for inference

GPU

Framework

None

Model used

Qwen3-Coder-30B-A3B-Instruct

Issue description

I tried running the Qwen3-Coder-30B-A3B-Instruct model on Intel Core Ultra 7 265, Ubuntu 25.10.

It worked on the CPU but failed on the GPU: During the GPU load attempt, system memory usage climbs until it fills the entire 96GB of RAM and exhausts nearly all of the 32GB swap file. I have attempted to mitigate this by applying various parameters to limit the KV cache size, but the memory leak/exhaustion persists regardless of these settings.

Step-by-step reproduction

I downloaded the optimized model pulled directly via:

/opt/openvino/ovms/bin/ovms --pull \
  --source_model "OpenVINO/Qwen3-Coder-30B-A3B-Instruct-int4-ov" \
  --model_repository_path /opt/openvino/models \
  --model_name Qwen3-Coder-30B-A3B-Instruct-int4-ov \
  --task text_generation

it worked with:

/opt/openvino/ovms/bin/ovms \
  --model_repository_path /opt/openvino/models \
  --model_name Qwen3-Coder-30B-A3B-Instruct-int4-ov \
  --task text_generation \
  --port 9001 \
  --rest_port 8000 \
  --target_device CPU

and failed (killed by Linux) with:

/opt/openvino/ovms/bin/ovms \
  --model_repository_path /opt/openvino/models \
  --model_name Qwen3-Coder-30B-A3B-Instruct-int4-ov \
  --task text_generation \
  --port 9001 \
  --rest_port 8000 \
  --target_device GPU

Relevant log output

Issue submission checklist

• I'm reporting an issue. It's not a question.
• I checked the problem with the documentation, FAQ, open issues, Stack Overflow, etc., and have not found a solution.
• There is reproducer code and related data files such as images, videos, models, etc.

[YuChern-Intel]

Duplicate case: #34187

[ikirsh]

Note that this is not the same model as in #34187.

This issue is regarding Qwen3-Coder-30B-A3B, which is a different model.

On #34187 Wan-Intel wrote: "On another note, you may open a new issue regarding the bug on ... as current thread will be focus on qwen3-30b-a3b."

So please note that it is not a duplicate case, considering your policy of one model per issue.

[YuChern-Intel]

I apologize for misreading the model name earlier.

However, Qwen3-Coder-30B-A3B model also can be run on GPU with OVMS on my side.

[ikirsh]

I have Intel Core Ultra 7 265 and I am trying to use the iGPU. No other GPU is installed.

# inxi -G
Graphics:
  Device-1: Intel Arrow Lake-S [Intel Graphics] driver: i915 v: kernel
  Display: unspecified server: X.Org v: 24.1.6 with: Xwayland v: 24.1.6
    driver: X: loaded: modesetting unloaded: fbdev,vesa dri: iris gpu: i915
    resolution: 1: 2560x1440~75Hz 2: 1920x1200~60Hz
  API: EGL v: 1.5 drivers: iris,swrast platforms: gbm,x11,surfaceless,device
  API: OpenGL v: 4.6 compat-v: 4.5 vendor: intel mesa
    v: 25.2.8-0ubuntu0.25.10.1 renderer: Mesa Intel Graphics (ARL)
  API: Vulkan v: 1.4.321 drivers: intel,llvmpipe surfaces: xcb,xlib
  Info: Tools: api: clinfo, eglinfo, glxinfo, vulkaninfo
    de: kscreen-console,kscreen-doctor gpu: gputop, intel_gpu_top, lsgpu
    wl: wayland-info x11: xdriinfo, xdpyinfo, xprop, xrandr

Can you try running this on your end using Intel Core Ultra 7 265 and no other GPU ?

It hangs/crashes right after 'Initializing Language Model Continuous Batching servable'.

Is there a flag to run this with more debug info so we can see what's failing?

[jgespino]

@ikirsh I was able to reproduce the issue on Ubuntu 24.04 using an Intel Core Ultra 2 265K processor. I’ll review this with the development team and update you once I have more information.

Ref. 182696