community pool / model answer

Qwen3.5-27B-GGUF

chat · unsloth/Qwen3.5-27B-GGUF

18.9 tok/s · model median · 12 runs in the model record.

Reference rig	Quantization	Median result	Basis	TTFT (ms)	Peak VRAM	Max context	Engines
GB10 Grace Blackwell 128GB	4-bit	11.1 tok/s	reported	-	- GB	262,144	llama.cpp
Minisforum UM790 Pro 64GB	4-bit	5 tok/s	reported	-	- GB	2,048	llama.cpp
Minisforum UM790 Pro 64GB	8-bit	2.8 tok/s	reported	-	- GB	2,048	llama.cpp
Multi-GPU 48GB ×2	8-bit	27.7 tok/s	reported	-	- GB	2,048	llama.cpp
Radeon 890M 32GB	4-bit	4.2 tok/s	reported	5,060	- GB	2,048	llama.cpp
Radeon AI Pro R9700 32GB	4-bit	26.7 tok/s	reported	466	- GB	2,048	llama.cpp
RTX 3090 24GB	4-bit	39.6 tok/s	reported	-	- GB	2,048	llama.cpp
RTX 4090 24GB	4-bit	45.4 tok/s	reported	-	- GB	2,048	llama.cpp
Ryzen AI Max 385 24GB	4-bit	4.1 tok/s	reported	5,060	- GB	2,048	llama.cpp

Context tested

The table above is the tested context for each exact rig and quantization cell. Empty fields remain empty rather than being filled with an estimate.