Can I run Mixtral-8x7B-Instruct-v0.1 on a NVIDIA H100 (80GB)?
Yes.Runs
You can run this with Q8 quantization. On a NVIDIA H100 (80GB) (80 GB VRAM), Mixtral-8x7B-Instruct-v0.1 needs about 57 GB (Q8) and should generate at roughly ~46 tok/s. Recommended: Run on your GPU with Q8.
mistralai/Mixtral-8x7B-Instruct-v0.1Why
- Full precision needs ~113 GB which exceeds your accelerator, but 8-bit (INT8 / Q8) needs only ~57 GB and fits.
How to run Mixtral-8x7B-Instruct-v0.1 on a H100 (80GB)
Run with Ollama (easiest)
ollama run mixtral:8x7bOr with llama.cpp
llama-cli -hf mistralai/Mixtral-8x7B-Instruct-v0.1 -p "Hello"Serve with vLLM (OpenAI-compatible API)
# Fast production serving on http://localhost:8000/v1
vllm serve mistralai/Mixtral-8x7B-Instruct-v0.1Python (Transformers)
from transformers import pipeline, BitsAndBytesConfig
# pip install transformers accelerate bitsandbytes
quant = BitsAndBytesConfig(load_in_8bit=True)
pipe = pipeline("text-generation", model="mistralai/Mixtral-8x7B-Instruct-v0.1", device_map="auto",
model_kwargs={"quantization_config": quant})
print(pipe("Hello", max_new_tokens=50))Want it to run comfortably at higher precision? The smallest GPU that runs Mixtral-8x7B-Instruct-v0.1 well is the NVIDIA H200 (141 GB).