Outmake Models
Models

Phi 3 Mini 128K

Microsoft · April 2024

Get Hugging Face

A 3.8B chat model from Microsoft.

Rating
2.1
Downloads
7.6K
Likes
42
Parameters
3.8B
Context
128K
Licence
mit

Runs great on M4 · 16 GB

The Q8_0 size from QuantFactory is the one to get: a 4.1 GB download that uses 6.0 GB of memory, and writes about 4.7 words a second.

Ratings

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

Text
1130
Coding
1155

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
43.9%

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 3 Mini 128K 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 QuantFactory/Phi-3-mini-128k-instruct-GGUF on Hugging Face. A smaller size is a smaller download and a little less accurate.

SizeDownloadMemory at 8KGet
Q2_K1.4 GB3.4 GBestimatedOpen
Q3_K_S1.7 GB3.7 GBestimatedOpen
Q3_K_M2.0 GB3.9 GBestimatedOpen
Q3_K_L2.1 GB4.1 GBestimatedOpen
Q4_02.2 GB4.1 GBestimatedOpen
Q4_K_S2.2 GB4.2 GBestimatedOpen
Q4_K_M2.4 GB4.4 GBestimatedOpen
Q4_12.4 GB4.4 GBestimatedOpen
Q5_02.6 GB4.6 GBestimatedOpen
Q5_K_S2.6 GB4.6 GBestimatedOpen
Q5_K_M2.8 GB4.8 GBestimatedOpen
Q5_12.9 GB4.8 GBestimatedOpen
Q6_K3.1 GB5.1 GBestimatedOpen
Q8_04.1 GB6.0 GBestimatedOpen

Run It Anywhere

Phi 3 Mini 128K is an open model, so any app that runs GGUF files can run it. These take the Q8_0 size from QuantFactory straight from Hugging Face.

  • LM StudioMac, Windows, LinuxOpen
  • JanMac, Windows, LinuxOpen
  • MstyMac, Windows, LinuxOpen
  • OllamaMac, Windows, Linuxollama run hf.co/QuantFactory/Phi-3-mini-128k-instruct-GGUF:Q8_0
  • llama.cppMac, Windows, Linuxllama-server -hf QuantFactory/Phi-3-mini-128k-instruct-GGUF:Q8_0
  • Docker Model RunnerMac, Windows, Linuxdocker model run hf.co/QuantFactory/Phi-3-mini-128k-instruct-GGUF:Q8_0
On a PC Graphics Card
Video memoryLargest size that fitsDownloadMemory at 8K
6 GBQ6_K3.1 GB5.1 GB
8 GBQ8_04.1 GB6.0 GB
10 GBQ8_04.1 GB6.0 GB
11 GBQ8_04.1 GB6.0 GB
12 GBQ8_04.1 GB6.0 GB
16 GBQ8_04.1 GB6.0 GB
20 GBQ8_04.1 GB6.0 GB
24 GBQ8_04.1 GB6.0 GB
32 GBQ8_04.1 GB6.0 GB
48 GBQ8_04.1 GB6.0 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 GBQ8_0Runs great6.0 GB6.8
M6 · 24 GBQ8_0Runs great7.9 GB6.8
M6 · 32 GBQ8_0Runs great12 GB6.8
M5 · 16 GBQ8_0Runs great6.0 GB6.2
M5 · 24 GBQ8_0Runs great7.9 GB6.2
M5 · 32 GBQ8_0Runs great12 GB6.2
M5 Pro · 24 GBQ8_0Runs great7.9 GB12
M5 Pro · 48 GBQ8_0Runs great19 GB12
M5 Pro · 64 GBQ8_0Runs great19 GB12
M5 Max · 36 GBQ8_0Runs great12 GB18
M5 Max · 48 GBQ8_0Runs great19 GB24
M5 Max · 64 GBQ8_0Runs great19 GB24
M5 Max · 128 GBQ8_0Runs great34 GB24
M5 Ultra · 96 GBQ8_0Runs great34 GB46
M5 Ultra · 256 GBQ8_0Runs great34 GB46
M5 Ultra · 512 GBQ8_0Runs great34 GB46
M4 · 16 GBQ8_0Runs great6.0 GB4.7
M4 · 24 GBQ8_0Runs great7.9 GB4.7
M4 · 32 GBQ8_0Runs great12 GB4.7
M4 Pro · 24 GBQ8_0Runs great7.9 GB10
M4 Pro · 48 GBQ8_0Runs great19 GB10
M4 Pro · 64 GBQ8_0Runs great19 GB10
M4 Max · 36 GBQ8_0Runs great12 GB15
M4 Max · 48 GBQ8_0Runs great19 GB19
M4 Max · 64 GBQ8_0Runs great19 GB19
M4 Max · 96 GBQ8_0Runs great34 GB15
M4 Max · 128 GBQ8_0Runs great34 GB19
M3 · 8 GBQ6_KRuns well5.1 GB3.7
M3 · 16 GBQ8_0Runs great6.0 GB4
M3 · 24 GBQ8_0Runs great7.9 GB4
M3 Pro · 18 GBQ8_0Runs great7.9 GB5.9
M3 Pro · 36 GBQ8_0Runs great12 GB5.9
M3 Max · 36 GBQ8_0Runs great12 GB11
M3 Max · 48 GBQ8_0Runs great19 GB15
M3 Max · 64 GBQ8_0Runs great19 GB15
M3 Max · 96 GBQ8_0Runs great34 GB11
M3 Max · 128 GBQ8_0Runs great34 GB15
M3 Ultra · 96 GBQ8_0Runs great34 GB27
M3 Ultra · 256 GBQ8_0Runs great34 GB27
M3 Ultra · 512 GBQ8_0Runs great34 GB27
M2 · 8 GBQ6_KRuns well5.1 GB3.7
M2 · 16 GBQ8_0Runs great6.0 GB4
M2 · 24 GBQ8_0Runs great7.9 GB4
M2 Pro · 16 GBQ8_0Runs great6.0 GB7.8
M2 Pro · 32 GBQ8_0Runs great12 GB7.8
M2 Max · 32 GBQ8_0Runs great12 GB15
M2 Max · 64 GBQ8_0Runs great19 GB15
M2 Max · 96 GBQ8_0Runs great34 GB15
M2 Ultra · 64 GBQ8_0Runs great19 GB27
M2 Ultra · 128 GBQ8_0Runs great34 GB27
M2 Ultra · 192 GBQ8_0Runs great34 GB27
M1 · 8 GBQ6_KRuns well5.1 GB2.5
M1 · 16 GBQ8_0Runs great6.0 GB2.7
M1 Pro · 16 GBQ8_0Runs great6.0 GB7.5
M1 Pro · 32 GBQ8_0Runs great12 GB7.5
M1 Max · 32 GBQ8_0Runs great12 GB14
M1 Max · 64 GBQ8_0Runs great19 GB14
M1 Ultra · 64 GBQ8_0Runs great19 GB25
M1 Ultra · 128 GBQ8_0Runs great34 GB25

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

Packaged by

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

Information

Maker
Microsoft
Released
April 2024
Parameters
3.8B
Longest chat
128K tokens
Reads pictures
No
Licence
mit
Downloads this month
7,611
Likes
42
Sizes
14, 1.4 to 4.1 GB
Kind
Maker's release

Badge for the model card

Shows on Phi 3 Mini 128K'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