Your very own tiny model is loading into this tab right now. It runs entirely locally on your device (no server, no cloud). Here's some context on the research that led to this model, and what I'm building next:
Small language models can have limited chat capabilities and be quite incoherent.
Try to chat with SmolLM2-135M for yourself - it is around the same size as my tiny model (it is actually a bit larger than mine): chat.webllm.ai.
If you are up for it, try chatting SmolLM2-360M as well on that site - it is over three times larger than my tiny model!
I am working on addressing current language model limitations, and publishing enjoyable and useful small models that run locally on computers and phones.
I am working on techniques towards general and recursive intelligence on the research side.
As far as new releases, I think upcoming releases will significantly improve performance without increasing model size, so stay tuned! I may also publish a tiny agent model.
I would love to hear from you. You can contact me through this Google Form.