community pool / model answer

Qwen3.6-35B-A3B

chat · unsloth/Qwen3.6-35B-A3B

57.6 tok/s · model median · 16 runs in the model record.

Reference rig	Quantization	Median result	Basis	TTFT (ms)	Peak VRAM	Max context	Engines
M3 Ultra 512GB	3-bit	84.6 tok/s	reported	224	- GB	8,192	lmstudio
Radeon RX 7900 XTX 24GB	3-bit	104.5 tok/s	reported	-	15.7 GB	640	llama.cpp
RTX 3090 24GB	4-bit	91.8 tok/s	reported	115	21.9 GB	32,768	llama.cpp
RTX 4060 Ti 16GB	3-bit	60.2 tok/s	reported	4,000	- GB	262,144	llama.cpp
RTX 4070 12GB	4-bit	57.4 tok/s	measured	68	11.6 GB	131,072	llama.cpp
Ryzen AI Max 395 128GB	4-bit	57.5 tok/s	measured	-	27.9 GB	131,072	llama.cpp
Ryzen AI Max 395 128GB	8-bit	52.8 tok/s	measured	-	40.4 GB	128,000	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.