Outmake Models
Models

Gemma 4 26B

Google · March 2026

GetHugging Face

Reads documents and screenshots, and ties for top LM Arena rating among Outmake's 24 GB Mac picks.

Rating
4.4
Downloads
2.8M
Parameters
26B, 3.8B active
Context
256K
Download
11 to 52 GB
Commercial use
Yes

Tight fit on M4 · 16 GB

Outmake picks the UD-IQ2_M size: a 11 GB download that uses 12 GB of memory, and writes about 18 words (25 tokens) a second.

About

Text and pictures can be mixed in any order within one question.

How it compares

Outmake estimates it is the 2nd fastest of the 8 chat models it recommends for a 24 GB Mac. Qwen3.6 35B is about 1.4 times as fast.

Among open models in LM Arena's public vote, it places 5th of 15 for reading pictures and 11th of 62 for text.

Of the chat models Outmake recommends for a 24 GB Mac, Qwen3.5 4B writes at about the same speed, by Outmake's estimate.

Limitations

It reads only about one minute of video.

What it knows stops in January 2025, and it may state wrong facts.

It can miss sarcasm, subtle nuance and figurative language.

Good to know

The A in its full name, 26B A4B, stands for the part of the model that is active at a time.

Sources: google/gemma-4-26B-A4B-it model card, Outmake's speed estimates, LM Arena and Hugging Face. Every line comes from one of these.

Ratings

4.4 out of 5, from how people voted on its answers.

Text
1438
Coding
1482
Vision
1257

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.

Good at

  • Pulling text from PDFs
  • Reading handwriting
  • Explaining charts and screens

Licence

Apache 2.0. Commercial use: yes. The maker's page has the full terms.

Gemma 4 26B runs well from a Mac with an M3 Pro and 18 GB of memory, at UD-IQ2_XXS, writing about 23 words (31 tokens) a second. 48 of 59 Mac configurations run it well.

Compare with

The nearest models in size, judged for the same Mac.

Sizes

SizeDownloadMemory at 8KMemory at 32KOpen
UD-Q3_K_MSmaller14 GB15 GB15 GBOpen
UD-Q4_K_XLBalanced18 GB19 GB19 GBOpen
Q8_0Best28 GB29 GB29 GBOpen
All 21 sizes
SizeDownloadMemory at 8KMemory at 32KOpen
UD-IQ2_XXS11 GB12 GBestimated12 GBOpen
UD-IQ2_M11 GB12 GBestimated12 GBOpen
UD-Q2_K_XL12 GB13 GBestimated13 GBOpen
UD-IQ3_S12 GB13 GBestimated14 GBOpen
UD-IQ3_XXS13 GB13 GBestimated14 GBOpen
UD-Q3_K_XL14 GB15 GBestimated15 GBOpen
UD-IQ4_XS15 GB16 GBestimated16 GBOpen
UD-IQ4_NL15 GB16 GBestimated16 GBOpen
UD-Q4_K_S18 GB18 GBestimated19 GBOpen
MXFP4_MOE18 GB19 GBestimated19 GBOpen
UD-Q4_K_M18 GB19 GBestimated19 GBOpen
UD-Q5_K_S20 GB21 GBestimated21 GBOpen
UD-Q5_K_M22 GB23 GBestimated23 GBOpen
UD-Q5_K_XL22 GB23 GBestimated23 GBOpen
UD-Q6_K24 GB25 GBestimated25 GBOpen
UD-Q6_K_XL24 GB25 GBestimated26 GBOpen
UD-Q8_K_XL29 GB30 GBestimated30 GBOpen
BF1652 GB52 GBestimated53 GBOpen

Memory is what llama.cpp's own calculator reported for each size when the catalog was built, with the app's settings. Smaller, Balanced and Best are the three sizes the store names.

Run It Anywhere

Gemma 4 26B is an open model, so any app that runs GGUF files can run it. These take the UD-Q4_K_XL size from unsloth straight from Hugging Face.

  • LM StudioMac, Windows, LinuxOpen
  • JanMac, Windows, LinuxOpen
  • MstyMac, Windows, LinuxOpen
  • OllamaMac, Windows, Linuxollama run hf.co/unsloth/gemma-4-26B-A4B-it-GGUF:UD-Q4_K_XL
  • llama.cppMac, Windows, Linuxllama-server -hf unsloth/gemma-4-26B-A4B-it-GGUF:UD-Q4_K_XL
  • Docker Model RunnerMac, Windows, Linuxdocker model run hf.co/unsloth/gemma-4-26B-A4B-it-GGUF:UD-Q4_K_XL
On a PC Graphics Card
Video memoryLargest size that fitsDownloadMemory at 8K
6 GBNone worth runningneeds 12 GB
8 GBNone worth runningneeds 12 GB
10 GBNone worth runningneeds 12 GB
11 GBNone worth runningneeds 12 GB
12 GBNone worth runningneeds 12 GB
16 GBUD-IQ4_NL15 GB16 GB
20 GBUD-Q4_K_XL18 GB19 GB
24 GBUD-Q5_K_XL22 GB23 GB
32 GBUD-Q8_K_XL29 GB30 GB
48 GBUD-Q8_K_XL29 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
MacSizeFitMemoryWritesReads
M6 · 16 GBUD-IQ2_MTight fit12 GB281158
M6 · 24 GBUD-Q3_K_MRuns well15 GB241158
M6 · 32 GBUD-Q3_K_MRuns great15 GB241158
M5 · 16 GBUD-IQ2_MTight fit12 GB26428
M5 · 24 GBUD-Q3_K_MRuns well15 GB22428
M5 · 32 GBUD-Q3_K_MRuns great15 GB22428
M5 Pro · 24 GBUD-Q3_K_MRuns well15 GB42855
M5 Pro · 48 GBUD-Q4_K_XLRuns great21 GB35855
M5 Pro · 64 GBQ8_0Runs great30 GB291069
M5 Max · 36 GBUD-Q3_K_MRuns great16 GB601711
M5 Max · 48 GBUD-Q4_K_XLRuns great21 GB642138
M5 Max · 64 GBQ8_0Runs great30 GB552138
M5 Max · 128 GBQ8_0Runs great30 GB552138
M5 Ultra · 96 GBQ8_0Runs great30 GB973422
M5 Ultra · 256 GBQ8_0Runs great30 GB974277
M5 Ultra · 512 GBQ8_0Runs great30 GB974277
M4 · 16 GBUD-IQ2_MTight fit12 GB18144
M4 · 24 GBUD-Q3_K_MRuns well15 GB16144
M4 · 32 GBUD-Q3_K_MRuns great15 GB16144
M4 Pro · 24 GBUD-Q3_K_MRuns well15 GB30287
M4 Pro · 48 GBUD-Q4_K_XLRuns great21 GB26287
M4 Pro · 64 GBQ8_0Runs great30 GB22287
M4 Max · 36 GBUD-Q3_K_MRuns great16 GB39575
M4 Max · 48 GBUD-Q4_K_XLRuns great21 GB41718
M4 Max · 64 GBQ8_0Runs great30 GB37718
M4 Max · 96 GBQ8_0Runs great30 GB30575
M4 Max · 128 GBQ8_0Runs great30 GB37718
M3 · 8 GBUD-Q3_K_MToo big16 GB
M3 · 16 GBUD-IQ2_MTight fit12 GB16119
M3 · 24 GBUD-Q3_K_MRuns well15 GB14119
M3 Pro · 18 GBUD-IQ2_XXSRuns well12 GB23208
M3 Pro · 36 GBUD-Q3_K_MRuns great16 GB20208
M3 Max · 36 GBUD-Q3_K_MRuns great16 GB34446
M3 Max · 48 GBUD-Q4_K_XLRuns great21 GB36595
M3 Max · 64 GBQ8_0Runs great30 GB31595
M3 Max · 96 GBQ8_0Runs great30 GB25446
M3 Max · 128 GBQ8_0Runs great30 GB31595
M3 Ultra · 96 GBQ8_0Runs great30 GB50892
M3 Ultra · 256 GBQ8_0Runs great30 GB50892
M3 Ultra · 512 GBQ8_0Runs great30 GB501190
M2 · 8 GBUD-Q3_K_MToo big16 GB
M2 · 16 GBUD-IQ2_MTight fit12 GB16117
M2 · 24 GBUD-Q3_K_MRuns well15 GB14117
M2 Pro · 16 GBUD-IQ2_MTight fit12 GB28234
M2 Pro · 32 GBUD-Q3_K_MRuns great15 GB25234
M2 Max · 32 GBUD-Q3_K_MRuns great15 GB41438
M2 Max · 64 GBQ8_0Runs great30 GB31438
M2 Max · 96 GBQ8_0Runs great30 GB31555
M2 Ultra · 64 GBQ8_0Runs great30 GB50876
M2 Ultra · 128 GBQ8_0Runs great30 GB50876
M2 Ultra · 192 GBQ8_0Runs great30 GB50876
M1 · 8 GBUD-Q3_K_MToo big16 GB
M1 · 16 GBUD-IQ2_MTight fit12 GB1195
M1 Pro · 16 GBUD-IQ2_MTight fit12 GB24189
M1 Pro · 32 GBUD-Q3_K_MRuns great15 GB22189
M1 Max · 32 GBUD-Q3_K_MRuns great15 GB33325
M1 Max · 64 GBQ8_0Runs great30 GB27325
M1 Ultra · 64 GBQ8_0Runs great30 GB40649
M1 Ultra · 128 GBQ8_0Runs great30 GB40649

Writes and reads are words a second (a token is about three quarters of a word), 4K tokens into a chat, estimated unless marked measured. The size is the one the store offers that Mac. Constants fitted October 11, 2026; how, and how wrong.

Badge for the model card

Shows on Gemma 4 26B's Hugging Face card which Macs run it well, linked to this page. It updates every morning.

Runs on Mac badge

Sizes and files are read from unsloth/gemma-4-26B-A4B-it-GGUF on Hugging Face and checked every morning. The rating is the model's score on the LM Arena text leaderboard (CC BY 4.0), where people vote between two answers without knowing which model wrote which. Catalog of November 4, 2026.

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