Instructions to use QuantTrio/gemma-4-31B-it-AWQ with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use QuantTrio/gemma-4-31B-it-AWQ with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="QuantTrio/gemma-4-31B-it-AWQ")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("QuantTrio/gemma-4-31B-it-AWQ")
model = AutoModelForMultimodalLM.from_pretrained("QuantTrio/gemma-4-31B-it-AWQ", device_map="auto")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
inputs = processor.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use QuantTrio/gemma-4-31B-it-AWQ with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "QuantTrio/gemma-4-31B-it-AWQ"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "QuantTrio/gemma-4-31B-it-AWQ",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/QuantTrio/gemma-4-31B-it-AWQ

SGLang

How to use QuantTrio/gemma-4-31B-it-AWQ with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "QuantTrio/gemma-4-31B-it-AWQ" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "QuantTrio/gemma-4-31B-it-AWQ",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "QuantTrio/gemma-4-31B-it-AWQ" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "QuantTrio/gemma-4-31B-it-AWQ",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Docker Model Runner
How to use QuantTrio/gemma-4-31B-it-AWQ with Docker Model Runner:
```
docker model run hf.co/QuantTrio/gemma-4-31B-it-AWQ
```

Request update model to google updates of july 2026

by bagwani - opened 15 days ago

Discussion

bagwani

15 days ago

Google made huge improvements to Gemma 4 tool-calling and chat accuracy, reliability + speed

Near complete list of Google's fixes/improvements:

Flash Attention 4: Uniform FA4 support on NVIDIA Hopper GPUs.
Speed gains: 25–70% higher prefill throughput and up to 31% lower time-to-first-token.
Chat formatting: Fixed null handling, input validation and unbalanced/missing turn tags.
Reasoning preservation: Corrected how thinking/reasoning content is retained and rendered.
Tool responses: Restored the assistant turn and thinking cue after tool outputs.
Tool calling: Improved execution accuracy, consistency and tool-call-only turn closure.
Continuation turns: Removed duplicate <turn|> tags and unwanted extra newlines.
Generation prompts: Reverted an add_generation_prompt regression and restored expected defaults.
Conversation history: Removed the obsolete APC thought primer and corrected historical-turn handling.
Template standardization: Added the canonical chat-template header and updated stale comments.
Vision controls: Default remains max_soft_tokens=280; use 1120 for sharper OCR and up to 2.51MP detail.
31B benchmark gains: BFCL +0.4, TB2 +4.5, Retail +3.1, Airline +2.0 and Telecom +10.1 points.

JunHowie

QuantTrio org 15 days ago

Thanks for the reminder. I’ve updated tokenizer_config.json from the upstream model repository.

JunHowie changed discussion status to closed 15 days ago

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment