Ask HN: What's the best LLM model that on a 24 GB VRAM GPU?
2•max93•1h ago
What’s the best model right now that outperforms Qwopus3.6-27B-v2-MTP-GGUF 8-bit on a 24 GB VRAM GPU? Looking for real reviews. I found 4 bit not usable in production.
Comments
roscas•30m ago
I would love to see Laguna.xs-2 on that, because it's very good on cpu only.
jr_isidore•8m ago
Good question. I was told in 2024 to get an RTX 3090 (24 GB VRAM), so I did, and nothing on HuggingFace was usable.
roscas•30m ago