OpenAssistant Llama — this is one of the best ChatGPT alternatives, but license is still a tricky issue. Hugginface has distributed XOR weights that seems to sidestep the license issue, but still a bit risky to use in enterprise settings. FastChat and Vicuna — this is a pretty decent model that looks to be next best in performance. Stable Vicuna — this the Stability AI’s variant of the famous Vicuna model that has Reinforcement learning with Human Feedback (RLHF) built-in. Open LLama — This is an Apache licensed variant of Llama that is just released. Cerebras-GPT — these are a family of Chinchilla based LLMs with varying sizes. OpenFlamingo — this is a multimodal LLM that is built based on DeepMind’s Flamingo and has a MIT license Flan — this is a fairly small model from Google, but can be finetuned for pretty cool applications.