Outmake Models
Models

Phi 4

Microsoft · December 2024

Get Hugging Face

A 15B chat model from Microsoft.

Rating
3.1
Downloads
72K
Likes
538
Parameters
15B
Context
16K
Licence
mit

Runs well on M4 · 16 GB

The Q4_1 size from microsoft is the one to get: a 9.3 GB download that uses 10 GB of memory, and writes about 4.9 words a second.

Ratings

3.1 out of 5, across people's blind votes on its answers and the tests below.

Text
1256
Coding
1306

LM Arena ratings of October 28, 2026 (CC BY 4.0): people vote between two answers without knowing which model wrote which. The bar ends at 1488, the best any open model has. The Arena tests the full model; a small size on a Mac answers a little worse.

KnowledgeMMLU-Pro
70.4%

MMLU-Pro: graduate-level questions in 14 subjects. Scores are the ones filed on the model's Hugging Face page or the board's own table, read 2026-11-05; some come from the maker.

Phi 4 runs well from a Mac with an M1 and 8 GB of memory. 59 of 59 Mac configurations run it well.

Compare with

The nearest models in size.

Sizes

From microsoft/phi-4-gguf on Hugging Face. A smaller size is a smaller download and a little less accurate.

SizeDownloadMemory at 8KGet
TQ1_03.6 GB4.7 GBestimatedOpen
TQ2_04.2 GB5.3 GBestimatedOpen
Q2_K5.5 GB6.6 GBestimatedOpen
IQ3_XXS6.1 GB7.2 GBestimatedOpen
IQ3_XS6.2 GB7.3 GBestimatedOpen
IQ3_S6.5 GB7.6 GBestimatedOpen
Q3_K_S6.5 GB7.6 GBestimatedOpen
IQ3_M6.9 GB8.0 GBestimatedOpen
Q3_K7.4 GB8.4 GBestimatedOpen
Q3_K_L7.9 GB9.0 GBestimatedOpen
IQ4_XS8.0 GB9.1 GBestimatedOpen
Q4_08.4 GB9.5 GBestimatedOpen
IQ4_NL8.4 GB9.5 GBestimatedOpen
Q4_K_S8.4 GB9.5 GBestimatedOpen
Q4_K9.1 GB10 GBestimatedOpen
Q4_19.3 GB10 GBestimatedOpen
Q5_010 GB11 GBestimatedOpen
Q5_K_S10 GB11 GBestimatedOpen
Q5_K11 GB12 GBestimatedOpen
Q5_111 GB12 GBestimatedOpen
Q6_K12 GB13 GBestimatedOpen
Q8_016 GB17 GBestimatedOpen
BF1629 GB30 GBestimatedOpen

Run It Anywhere

Phi 4 is an open model, so any app that runs GGUF files can run it. These take the Q4_1 size from microsoft straight from Hugging Face.

  • LM StudioMac, Windows, LinuxOpen
  • JanMac, Windows, LinuxOpen
  • MstyMac, Windows, LinuxOpen
  • OllamaMac, Windows, Linuxollama run hf.co/microsoft/phi-4-gguf:Q4_1
  • llama.cppMac, Windows, Linuxllama-server -hf microsoft/phi-4-gguf:Q4_1
  • Docker Model RunnerMac, Windows, Linuxdocker model run hf.co/microsoft/phi-4-gguf:Q4_1
On a PC Graphics Card
Video memoryLargest size that fitsDownloadMemory at 8K
6 GBNone worth runningneeds 7.2 GB
8 GBQ3_K_S6.5 GB7.6 GB
10 GBQ4_K_S8.4 GB9.5 GB
11 GBQ4_19.3 GB10 GB
12 GBQ5_K11 GB12 GB
16 GBQ6_K12 GB13 GB
20 GBQ8_016 GB17 GB
24 GBQ8_016 GB17 GB
32 GBBF1629 GB30 GB
48 GBBF1629 GB30 GB

The largest size, of those that keep three bits a weight or more, whose memory at an 8K chat is under 92% of the card's; the rest is the desktop's. Estimated from the files, the same way as for a Mac; speed on a PC is not estimated yet. A size that doesn't fit the card still runs, slower, with part of it in ordinary memory.

On every Mac
MacSizeFitMemoryWrites
M6 · 16 GBQ4_1Runs well10 GB6.9
M6 · 24 GBQ5_K_MRuns great11 GB5.7
M6 · 32 GBQ6_KRuns great14 GB5.4
M5 · 16 GBQ4_1Runs well10 GB6.3
M5 · 24 GBQ5_K_MRuns great11 GB5.1
M5 · 32 GBQ6_KRuns great14 GB4.9
M5 Pro · 24 GBQ5_K_MRuns great11 GB10
M5 Pro · 48 GBQ8_0Runs great18 GB9.3
M5 Pro · 64 GBQ8_0Runs great18 GB9.3
M5 Max · 36 GBQ8_0Runs great17 GB14
M5 Max · 48 GBQ8_0Runs great18 GB18
M5 Max · 64 GBQ8_0Runs great18 GB18
M5 Max · 128 GBQ8_0Runs great18 GB18
M5 Ultra · 96 GBQ8_0Runs great18 GB35
M5 Ultra · 256 GBQ8_0Runs great18 GB35
M5 Ultra · 512 GBQ8_0Runs great18 GB35
M4 · 16 GBQ4_1Runs well10 GB4.9
M4 · 24 GBQ5_K_MRuns great11 GB4
M4 · 32 GBQ6_KRuns great14 GB3.7
M4 Pro · 24 GBQ5_K_MRuns great11 GB8.8
M4 Pro · 48 GBQ8_0Runs great18 GB7.8
M4 Pro · 64 GBQ8_0Runs great18 GB7.8
M4 Max · 36 GBQ8_0Runs great17 GB11
M4 Max · 48 GBQ8_0Runs great18 GB15
M4 Max · 64 GBQ8_0Runs great18 GB15
M4 Max · 96 GBQ8_0Runs great18 GB11
M4 Max · 128 GBQ8_0Runs great18 GB15
M3 · 8 GBTQ2_0Runs well5.3 GB6
M3 · 16 GBQ4_1Runs well10 GB4.1
M3 · 24 GBQ5_K_MRuns great11 GB3.4
M3 Pro · 18 GBQ3_KRuns great8.4 GB6.1
M3 Pro · 36 GBQ8_0Runs great17 GB4.4
M3 Max · 36 GBQ8_0Runs great17 GB8.6
M3 Max · 48 GBQ8_0Runs great18 GB11
M3 Max · 64 GBQ8_0Runs great18 GB11
M3 Max · 96 GBQ8_0Runs great18 GB8.6
M3 Max · 128 GBQ8_0Runs great18 GB11
M3 Ultra · 96 GBQ8_0Runs great18 GB21
M3 Ultra · 256 GBQ8_0Runs great18 GB21
M3 Ultra · 512 GBQ8_0Runs great18 GB21
M2 · 8 GBTQ2_0Runs well5.3 GB6
M2 · 16 GBQ4_1Runs well10 GB4.1
M2 · 24 GBQ5_K_MRuns great11 GB3.4
M2 Pro · 16 GBQ4_1Runs well10 GB8
M2 Pro · 32 GBQ6_KRuns great14 GB6.1
M2 Max · 32 GBQ6_KRuns great14 GB12
M2 Max · 64 GBQ8_0Runs great18 GB11
M2 Max · 96 GBQ8_0Runs great18 GB11
M2 Ultra · 64 GBQ8_0Runs great18 GB21
M2 Ultra · 128 GBQ8_0Runs great18 GB21
M2 Ultra · 192 GBQ8_0Runs great18 GB21
M1 · 8 GBTQ2_0Runs well5.3 GB4
M1 · 16 GBQ4_1Runs well10 GB2.8
M1 Pro · 16 GBQ4_1Runs well10 GB7.7
M1 Pro · 32 GBQ6_KRuns great14 GB5.9
M1 Max · 32 GBQ6_KRuns great14 GB11
M1 Max · 64 GBQ8_0Runs great18 GB11
M1 Ultra · 64 GBQ8_0Runs great18 GB19
M1 Ultra · 128 GBQ8_0Runs great18 GB19

Writes is words a second, estimated from the Mac's memory speed and the model's size; how, and how wrong.

Packaged by

6 repositories on Hugging Face carry this model for llama.cpp. Outmake reads them all as one model.

Information

Maker
Microsoft
Released
December 2024
Parameters
15B
Longest chat
16K tokens
Reads pictures
No
Licence
mit
Downloads this month
71,848
Likes
538
Sizes
23, 3.6 to 29 GB
Kind
Maker's release

Badge for the model card

Shows on Phi 4's Hugging Face card which Macs run it well, linked to this page. It updates every morning.

Runs on Mac badge

Files, downloads and likes read from Hugging Face on November 5, 2026 and refreshed every morning. Memory and speed are estimates until someone running Outmake measures this model on that Mac.

Your Computer

Every model is judged for this Mac. Both are under About This Mac, in the Apple menu.

Every Mac · Every graphics card

Civitai

Open in OutmakeHugging Face

New to Outmake? Download the free beta