An open-source family of large language models developed by Alibaba Cloud, featuring scalable model sizes and versions released under permissive licenses like Apache 2.0 .
An open-source reproduction of Meta AI’s LLaMA models, offering permissively licensed weights in 3B, 7B, and 13B parameter sizes compatible with both PyTorch and JAX.
A fully open-source LLM framework from the Allen Institute for AI, providing model weights, training code, data, and evaluation tools in a transparent package.
A fully-managed vector database service powered by Milvus, designed for enterprise-grade AI applications at any scale.
A fully open-source, 1B-parameter Chinese-centric language model trained from scratch, with complete access to code, data, checkpoints, and logs under the Apache 2.0 license.
A compact, open-source LLaMA-based language model (~1.1 B parameters) pretrained on trillions of tokens under the Apache 2.0 license.
A fully open-source large language model suite offering reproducible training, open weights, and instruction-tuned variants under the Apache 2.0 license.
Qdrant is a self-hosted vector database and similarity search engine written in Rust, optimized for AI applications and semantic search. Apache 2.0 licensed.
GPT-NeoX is an open source LLM training framework from EleutherAI for training and fine-tuning large language models on GPU clusters at research scale. Apache 2.0.
Open-source, declarative orchestration platform to build, run, and monitor data pipelines and workflows as code or via UI.
LLM Foundry is an open source LLM fine-tuning framework by MosaicML for training and instruction-tuning Llama, Mistral, and DBRX on custom data with efficient GPU utilization. Apache 2.0.
