MPT Foundation Series
Commercially usable open-source LLMs.
Optimized for fast training and inference.
Why the ML Community Loves MPT
Open Source
with full data transparency
Licensed
for the possibility of commercial use
Customizable
on your data for your use case
Fast Training + Inference
for cost effective deployments
8k+ Context Length
to understand long documents
Optimized Code
for pretraining, finetuning, evaluation
Testimonial
“Using the MosaicML platform, we were able to train and deploy our Ghostwriter 2.7B LLM for code generation with our own data within a week and achieve leading results.”
Meet the MPT Family
A new standard for open-source, commercially usable LLMs. Train, finetune, and deploy your own private models.
Train and Deploy Custom LLMs with MosaicML
Unleash the power of generative AI and maximize the potential of your data with our purpose-built platform.
MosaicML
Finetuning
Customize models on your data for your specific use cases
MosaicML
Pretraining
Create your own domain-specific LLMs from scratch for maximum customization & security
Testimonial
“The foundation series by MosaicML, including MPT-7B/30B (and an efficient training repo), makes high-quality pre-trained language models available to anyone for commercial use.”
LLM Foundry
Use our optimized codebase for training, finetuning, evaluating, and deploying LLMs.
Our codebase was built with Composer and integrates with the MosaicML platform.
Easy to use, efficient, and flexible, this codebase is designed to enable rapid experimentation with the latest techniques.
Model Gauntlet
The Model Gauntlet is our standardized method of evaluating LLMs in a holistic manner.
It encompasses 34 different benchmarks collected from a variety of sources, and organized into several broad categories of competencies.
Check out our leaderboard to compare eval scores across popular open-source LLMs!
Testimonial
“MPT-30B is an open-source and commercially licensed decoder-based LLM that is more powerful than GPT-3-175B.”
In the Community
Built with MPT:
- ReplitLM: replit-code-v1-3b is a 2.7B Causal Language Model focused on Code Completion. The model has been trained on a subset of the Stack Dedup v1.2 dataset covering 20 languages such as Java, Python, and C++
- LLaVa-MPT: Visual instruction tuning to get MPT multimodal capabilities
- ggml: Optimized MPT version for efficient inference on consumer hardware
- GPT4All: locally running chat system, now with MPT support!
- Q8MPT-Chat: 8-bit optimized MPT for CPU by our friends at Intel
- Gorilla: Large Language Model Connected with Massive APIs
- Tutorial: Using MPT-7B with Langchain and Hugging Face
Start your LLM journey with MosaicML
