Release of an 8B model?
This is awesome guys!!! You really have outdone yourselves. Can we expect an 8 billion parameter version in the near future as well? 😋
I suppose so,
It was on the release comments. I hope soon. Right now qwen 2.5 7b is the best small language for laptops.
And llama 3.2 3B for phones with Snapdragon cpus.
If the 8B llama has the same quality increase as the 405B->70B jump, then we will finally be able to run the very first capable LLM on consumer devices with normal GPUs (12gb vram)
The 70B is in my tests as good as the 405B model, and at reasoning it even excels 405B.
I wonder how they do that.
In all cases the 8B model might change everything, if we can run a capable model on consumer grade devices. No more need to go to chatgpt for normal tasks. This would offload all the server parks, if consumers run locally.