Qwen3 beat gpt-oss with just 0.6B with good quality enough to be usable
Qwen3 beat gpt-oss with just 0.6B with good quality enough to be usable.
OpenAi must make family of it not just two big models?
I dreamt to see something like that:
gpt-oss-0.6b
gpt-oss-1.5b
gpt-oss-3b
gpt-oss-7b
before I see two big models like:
gpt-oss-20b
gpt-oss-120b
These models are both capable of running on 8GB VRAM. That's one of the only things they did well, and it's done shockingly well.
These models are both capable of running on 8GB VRAM. That's one of the only things they did well, and it's done shockingly well.
I have to say I'm using Samsung M33 5G with 8GB ram and 8GB vram and I can run 2B parameters with slow speed so, what i need actually is more effective models not just make it big to be clever? gemma3 1B parameters is clever more then every single model that i tested already.
What i mean it's possible to make it smaller.
Free models are made to run on a phone or old tablet. I hate how censored this model is, but your still asking too much. Running very capable AI is going to take some kind of investment on your part.