Model optimization
Mistral compression
Mistral models are popular for self-hosted deployments due to their efficiency. Compression further optimizes inference by reducing prompt prefill time.
Mistral with compression
from supercompress import Compressor
from mistralai import Mistral
comp = Compressor()
client = Mistral(api_key="...")
def chat_with_compression(messages, query):
history = "\n".join(m["content"] for m in messages[:-1])
compressed = comp.compress(history, query)
messages[-1]["content"] = compressed.compressed_text + "\n" + query
return client.chat.complete(model="mistral-large", messages=messages)Frequently asked questions
Does compression work with Mistral's function calling?
Yes. Function definitions are preserved. Only conversation history is compressed.
Can I use it with self-hosted Mistral?
Yes. The compressor runs locally and compresses before sending to your self-hosted instance.
Build with less context
Put compression in front of your next LLM call.
Use the hosted API or run SuperCompress locally. Keep the evidence, drop the token waste, and measure the savings before it reaches your model.