Reddit users discuss coding models for 64GB VRAM

One user recommends an unsloth Qwen 3.5 122b-a10b at 30 tok/sec with 100k context. The thread explores alternatives for local LLM coding setups.