community pool / model answer

Phi-4-mini-reasoning-GGUF

chat · unsloth/Phi-4-mini-reasoning-GGUF

74.3 tok/s · model median · 47 runs in the model record.

Reference rig	Quantization	Median result	Basis	TTFT (ms)	Peak VRAM	Max context	Engines
RTX 3060 12GB	4-bit	65.2 tok/s	measured	781	- GB	131,072	llama.cpp
RTX 3060 12GB ×2	4-bit	73.3 tok/s	measured	1,838	8.6 GB	131,072	llama.cpp
RTX 3090 24GB	4-bit	121.6 tok/s	measured	1,403	- GB	131,072	llama.cpp
Ryzen AI Max 395 64GB	4-bit	73.9 tok/s	reported	515	3.9 GB	8,192	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.