Custom LLMs without sharing compute

Running large language models (LLMs) on Salad provides a convenient, cost-effective solution to deploy various applications without worrying about infrastructure. Host your personal, proprietary LLMs without sharing compute, protecting your data/prompts from being trained on.

LLM inference benchmark on SaladCloud

Have questions about SCE for your workload?

Discuss your use case with our team to see if SCE is fit for you.

Run popular models or bring your own

LLM inference hosting cloud - Salad

LLM Inference Hosting

As more LLMs are optimized to serve inference on GPUs with lower vRAM, Salad's network of RTX/GTX GPUs with the lowest GPU prices can save you thousands of dollars - offering enhanced efficiency and reduced costs.

per Million tokens
Average Text-generation inference (TGI) cost for Mistral 7B, Falcon-7B and CodeLlama on Salad.
Starting price
Deploy your own LLM with Ollama and HuggingChat UI on SaladCloud infrastructure.

Custom LLMs with UI

Deploy custom LLMs to 1000s of GPUs at the lowest prices, scaling easily at an affordable cost. Bring your models to life with a user-friendly UI, such as HuggingFace ChatUI.

starting price
Save 50% or more for your custom LLMs with self-managed, open-source models on Salad.
Speech to text - Automatic speech recognition with Whisper large on Salad GPU cloud
LLM enterprise chatbot with Ollama and LangChain

Enterprise Chatbots

Run your own Retriever-Augmented Generative (RAG) models with LangChain and Ollama to query your enterprise's own data. Deploy/scale popular models in a customizable, cost-effective way with Text-Generation Inference (TGI).

cost per hour
Run 7 Billion parameter models for just $0.22/hour on SaladCloud, a low-cost solution for custom GPT models.
cost per hour
Ensure seamless integration with TGI and deliver optimum performance for your enterprise chatbots