Hacker News Reader

Thank you. Which is currently the most capable version running reasonably fast on a 3090 (24GB of VRAM)?

The Llama distilled version Q4_K_M should be reasonably fast and good!!