models · microsoft

Phi-4

14B small language model trained on curated and synthetic data. Punches far above its weight on reasoning and STEM.

What this registry has checked

Quarantine. We hold the publisher’s own published hash for these weights and have recorded it against this entry. We have not downloaded and hashed the weights ourselves, so the hash is the publisher’s claim, not our measurement.

Weight file SHA-256	e956d92aecc2eefed264bb1da3fe5e73ba9c6a4db3810c78c32c6c8e42dde09a
Method	hf-published-sha256

Re-check this yourself → Source → Full record →

What is recorded

Publisher	microsoft
Origin	United States
Licence	MIT
Parameters	14,659,507,200 (14.7B)
Context window	16K
Modality	text->text
Architecture	phi3
Layers	40
Hidden size	5120
Attention	GQA (40 heads, 10 KV)
Native precision	BF16
Pulls recorded here	540,020
Downloads reported upstream	695,566

Reported scores

MMLU-Pro	71.4
GPQA Diamond	57.5
AIME 2025	18
MATH-500	81
LiveCodeBench	23.1
IFBench	23.5
Humanity's Last Exam	3.8

Reported by a third party and recorded here, not re-run by us. Source: artificialanalysis.ai/api/v2.

Measured on real hardware

None of these runs are ours. Each figure is the reporter’s, linked to their own evidence, under their own trust label. The number shown is single-stream decode — one user, concurrency 1 — because aggregate throughput across many users is a different quantity and reads as far faster than anyone will see.

Reported to local-ai-registry

Hardware	Engine	Build measured	Decode, 1 stream	Label	Reported by
GeForce RTX 3080	llama.cpp	Q4_K_M	77.0 tok/s @ 2,048	candidate	localmaxxing
GeForce RTX 3060	ollama	Q4_K_M	36.0 tok/s @ 2,048	candidate	localmaxxing
GeForce RTX 3060 ×2 devices	llama.cpp	Q4_K_M	34.2 tok/s @ 16,384	candidate	localmaxxing
Measured outside 2K–32K context — decode speed depends heavily on context, so these are not comparable with the rows above
GeForce RTX 3090	llama.cpp	Q4_K_M	47.1 tok/s @ 512	candidate	localmaxxing
GeForce RTX 3060 ×2 devices	llama.cpp	Q4_K_M	35.4 tok/s @ 512	candidate	localmaxxing
GeForce RTX 3060	llama.cpp	Q4_K_M	34.5 tok/s @ 512	candidate	localmaxxing

From local-ai-registry (MIT), commit 124959d: 6 runs across 3 machines, 0 validated — model revision and runtime pinned, launch accepted — and 6 candidate, their label for useful evidence without a reproducible-launch promise.

Elsewhere in the registry

All models → Browse with filters → What will run on my hardware? →

Registry record as of 2026-06-11. Identity, licence and parameter count are properties of a release and do not move; tier can, and the verify link above re-checks it against the live registry rather than this page.