family member

Phi 4 BASE

2 builds of Phi 4 BASE measured on real machines — different packagers, precisions and engines, each a different file with its own numbers. The best each engine proved is first; every build has its own page with the machines and signatures behind it.

on ollama · 1 build

no context proved on this engine yet

3.6 tok/s peak, by microsoft BF16

on vllm · 1 build

proved 32K context proved by redhatai

proved by at least one build: Structured output · Sustained output · Writes code · Reasons about code · Synthesises sources · Follows instructions

build	publisher	precision	engine	whose work	context proved	best tok/s	machines
RedHatAI/phi-4-quantized.w8a8	redhatai	—	vllm	—	32,768	—	1
hf.co/microsoft/phi-4-gguf:BF16	microsoft	BF16	ollama	vendor	—	3.6	1