Instructions to use majentik/Mistral-Small-4-119B-RotorQuant with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use majentik/Mistral-Small-4-119B-RotorQuant with Transformers:

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

pipe = pipeline("image-text-to-text", model="majentik/Mistral-Small-4-119B-RotorQuant")

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("majentik/Mistral-Small-4-119B-RotorQuant", dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use majentik/Mistral-Small-4-119B-RotorQuant with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "majentik/Mistral-Small-4-119B-RotorQuant"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "majentik/Mistral-Small-4-119B-RotorQuant",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/majentik/Mistral-Small-4-119B-RotorQuant

SGLang

How to use majentik/Mistral-Small-4-119B-RotorQuant 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 "majentik/Mistral-Small-4-119B-RotorQuant" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "majentik/Mistral-Small-4-119B-RotorQuant",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "majentik/Mistral-Small-4-119B-RotorQuant" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "majentik/Mistral-Small-4-119B-RotorQuant",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use majentik/Mistral-Small-4-119B-RotorQuant with Docker Model Runner:
```
docker model run hf.co/majentik/Mistral-Small-4-119B-RotorQuant
```

majentik commited on Apr 21

Commit

e0a58a5

verified ·

1 Parent(s): 7149f51

chore(card): add hardware compatibility section

Browse files

Files changed (1) hide show

README.md +11 -8

README.md CHANGED Viewed

@@ -2,16 +2,13 @@
 license: apache-2.0
 base_model: mistralai/Mistral-Small-4-119B-2603
 tags:
-- rotorquant
-- kv-cache-quantization
-- mistral
-- moe
-- quantized
 library_name: transformers
 pipeline_tag: image-text-to-text
-language:
-- en
-inference: false
 ---
 # Mistral-Small-4-119B-RotorQuant
@@ -20,6 +17,12 @@ inference: false
 This is a **documentation repository** that explains how to combine Mistral-Small-4-119B's weights with RotorQuant inference-time KV cache compression. No weights are stored here — use the base model directly and apply RotorQuant via the Python package or llama.cpp fork.
 ## What is this?
 KV cache compression reduces the memory used by the attention cache during inference. Unlike weight quantization (which is baked into the GGUF/MLX file), KV cache compression is applied at runtime — so the same base weights can be used with or without compression.

 license: apache-2.0
 base_model: mistralai/Mistral-Small-4-119B-2603
 tags:
+  - rotorquant
+  - kv-cache-quantization
+  - mistral
+  - moe
+  - quantized
 library_name: transformers
 pipeline_tag: image-text-to-text
 ---
 # Mistral-Small-4-119B-RotorQuant
 This is a **documentation repository** that explains how to combine Mistral-Small-4-119B's weights with RotorQuant inference-time KV cache compression. No weights are stored here — use the base model directly and apply RotorQuant via the Python package or llama.cpp fork.
+## Hardware compatibility
+| Device | VRAM / RAM | Recommendation |
+| --- | --- | --- |
+| Any host that runs the base model | baseline + runtime savings | RotorQuant/TurboQuant is a KV-cache runtime modifier; pair with any weight variant |
 ## What is this?
 KV cache compression reduces the memory used by the attention cache during inference. Unlike weight quantization (which is baked into the GGUF/MLX file), KV cache compression is applied at runtime — so the same base weights can be used with or without compression.