Here is a list of available LLMs (large language model) and VLMs (vision language models) which are open-source and potentially one can run them on standard computers.

Please note that for using these models you do not necessarily need to download them directly or use the official instructions for driving them, but can use tools such as Ollama (local), HuggingFace (local and remote), CoLab (remote), ... . In some of these tools there is a facilitated way of using many of these LLMs/VLMs hence you do not need to know a link to the original repository (which I put anyway).

Last update: 2025-01-29 (things change very fast!)

LLMS

LLama (from Meta) https://github.com/meta-llama/llama-models
DeepSeek R (in various editions) https://github.com/deepseek-ai/DeepSeek-R1
Mistral series (Mistral, Codestral, Pixtral, ...) https://huggingface.co/mistralai
GPT-SW3 (LLM for Swedish language and more) https://huggingface.co/AI-Sweden-Models/gpt-sw3-356m
Chronos-Bolt (time-series model, from Amazon) https://github.com/amazon-science/chronos-forecasting?tab=readme-ov-file

VLMS

LAVIS (also called BLIP): Lightweight and flexible, suitable for multimodal applications. (https://github.com/salesforce/LAVIS?tab=readme-ov-file)
MiniGPT-4: A smaller-scale vision-language model that runs on mid-range GPUs. (https://github.com/Vision-CAIR/MiniGPT-)
OpenAI CLIP: A model for connecting text and images. ( https://github.com/openai/CLIP)
OpenFlamingo: An open-source version of Flamingo optimized for accessible hardware.( https://github.com/mlfoundations/open_flamingo )
Janus Pro: A promising, especially with its multimodal capabilities. If computational resources are limited. (https://github.com/deepseek-ai/Janus)