LOCAL AIregistry
‹ all GPUs

RTX 4000 Ada

NVIDIA · 20 GB · 360 GB/s
The recipe
Recommended
Qwen3.8-27B
EXL3 4 bpwTabbyAPI128K context

Best for Coding agents, long documents, everyday assistant work.

37 tok/sdecode
–prefill
128Kcontext
ThinksToolsLong context✓ Vision
Earlier check · Sep 23, 2026