selfhost.directory

The index of self-hostable projects — ranked, categorized and served with live updates.
Own your data and stop giving it to big tech.

google logo

Can the Arc Pro B60 run gemma-4-26B-A4B-it?

gemma-4-26B-A4B-it·25.8B (4B active)·Arc Pro B60·24 GB VRAM

Saved in this browser, used on every model page.
Yes

Yes. The Arc Pro B60 runs gemma-4-26B-A4B-it fully in VRAM at Q5_K_M (about 20 GB of 24 GB), at roughly 88 tokens/s.

best quant
Q5_K_M
needs
20 GB
usable
24 GB
est. speed
~88 tok/s

Estimates for an 8K context; offload assumes 32 GB of system RAM.Try other contexts on the gemma-4-26B-A4B-it page →

Every quantisation of gemma-4-26B-A4B-it on the Arc Pro B60

QuantQualityMemoryFitSpeed
Q2_K Heavy loss11 GBfits~191 tok/s
Q3_K_M Noticeable loss15 GBfits~128 tok/s
Q4_K_M Small loss18 GBfits~103 tok/s
Q5_K_M pickNear-lossless20 GBfits~88 tok/s
Q6_K Near-lossless23 GBtight~76 tok/s
Q8_0 Lossless29 GBoffload~28 tok/s
FP16 / BF16 Lossless52 GBtoo big—

Speed is a bandwidth-based estimate for generating text (456 GB/s); prompt processing and long contexts are slower.

Measured speed on the Arc Pro B60

No community reports for this pair yet.

Report your own speed →

Run gemma-4-26B-A4B-it on the Arc Pro B60

Using Q5_K_M, the best fit for Arc Pro B60.

Ollama · From the Ollama librarydocs
ollama run gemma4:26b

The library build picks its own default quantisation. Ollama’s ROCm/CUDA builds don’t cover this GPU. Recent Ollama releases use Vulkan when its Vulkan backend and driver are installed (older ones needed OLLAMA_VULKAN=1); llama.cpp or LM Studio with Vulkan are the safer bet.

Ollama · UD-Q5_K_M straight from Hugging Facedocs
ollama run hf.co/unsloth/gemma-4-26B-A4B-it-GGUF:UD-Q5_K_M

Ollama’s ROCm/CUDA builds don’t cover this GPU. Recent Ollama releases use Vulkan when its Vulkan backend and driver are installed (older ones needed OLLAMA_VULKAN=1); llama.cpp or LM Studio with Vulkan are the safer bet.

llama.cpp · OpenAI-compatible server with a web UIdocs
llama-server -hf unsloth/gemma-4-26B-A4B-it-GGUF:UD-Q5_K_M

Needs a Vulkan build of llama.cpp. Downloads the GGUF on first run.

llama.cpp · Docker with Vulkandocs
docker run -d --name llama --device /dev/dri -p 8080:8080 \
  -v ~/.cache/llama.cpp:/root/.cache/llama.cpp \
  ghcr.io/ggml-org/llama.cpp:server-vulkan \
  -hf unsloth/gemma-4-26B-A4B-it-GGUF:UD-Q5_K_M --host 0.0.0.0 --port 8080 -ngl 99

Server and web UI on port 8080.

LM Studio · Download with the lms CLIdocs
lms get unsloth/gemma-4-26B-A4B-it-GGUF@UD-Q5_K_M

Then load it from the LM Studio app or with `lms load`.

Transformers · Ask about an imagedocs
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="google/gemma-4-26B-A4B-it", dtype="auto", device_map="auto")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/bee.jpg"},
            {"type": "text", "text": "Describe this image."},
        ],
    }
]
out = pipe(text=messages, max_new_tokens=256)
print(out[0]["generated_text"][-1]["content"])

Vision models that fit the Arc Pro B60 comfortably

All that fit →
ModelParamsBest quantMemorySpeedLicence
27.8BQ5_K_M≈22 GBFits in memory~13 tok/sCommercial OK
27.8BQ5_K_M≈22 GBFits in memory~13 tok/sCommercial OK
27.4BQ4_K_M≈21 GBFits in memory~15 tok/sConditions
27.8BQ5_K_M≈22 GBFits in memory~13 tok/sCommercial OK
meta-models logoMuse-Glimmer-30Bmeta-models
29.8BQ5_K_M≈22 GBFits in memory~12 tok/sCommercial OK
36BA3BQ4_K_M≈22 GBFits in memory~138 tok/sCommercial OK

Other GPUs that run gemma-4-26B-A4B-it

Frequently asked questions

Can the Arc Pro B60 run gemma-4-26B-A4B-it?
Yes. The Arc Pro B60 runs gemma-4-26B-A4B-it fully in VRAM at Q5_K_M (about 20 GB of 24 GB), at roughly 88 tokens/s.
How fast does gemma-4-26B-A4B-it run on the Arc Pro B60?
An estimate from memory bandwidth gives about 88 tokens/s at Q5_K_M.
Which quantisation of gemma-4-26B-A4B-it should I use on the Arc Pro B60?
Q5_K_M: the largest one that fits. Expected quality: near-lossless compared with the full-precision weights.

Memory and speed are estimates from the model's size and the card's memory bandwidth; real numbers depend on the runtime, drivers and context length. Community figures are medians of user reports.