RTX 5090

NVIDIA · 32 GB · 1792 GB/s

Top 3 recipes

Recommended

Qwen3.8-27B

EXL3 5 bpwTabbyAPI256K context

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

155 tok/sdecode

1,407 tok/sprefill

256Kcontext

✓ Thinks✓ Tools✓ Long context✓ Vision

Tested on this card · Sep 25, 2026

Fastest

Qwen3.5-9B

EXL3 8 bpwTabbyAPI256K context

Best for Fast answers and light agent work on 8–12 GB cards.

206 tok/sdecode

3,164 tok/sprefill

256Kcontext

✓ Thinks✓ Tools✓ Long context

Tested on this card · Sep 26, 2026

Another family

Gemma 4 26B A4B

EXL3 6.1 bpwTabbyAPI128K context

Best for Fast multimodal work on 16–24 GB cards.

173 tok/sdecode

4,483 tok/sprefill

128Kcontext

✓ Thinks✓ Tools✓ Long context✓ Vision

Tested on this card · Sep 26, 2026