Hugging FaceFor reading photos, documents and speech, and the most downloaded 8 GB Mac pick on Hugging Face last month.
- Rating, estimated
- 3.3
- Downloads
- 2.8M
- Parameters
- 8B
- Context
- 128K
- Download
- 4.5 to 16 GB
- Commercial use
- Yes
Runs great on M4 · 16 GB
Outmake picks the UD-Q5_K_XL size: a 7.6 GB download that uses 7.6 GB of memory, and writes about 8.9 words (12 tokens) a second.
About
It handles more than 35 languages out of the box and was trained on more than 140.
How it compares
Of the 7 chat models Outmake recommends for an 8 GB Mac, it is the 6th fastest by Outmake's estimate.
Outmake estimates K2-Horizon 3.7B writes about 1.2 times as fast on an 8 GB Mac, but it can't read pictures.
Outmake estimates Qwen3.5 0.8B writes about 6.9 times as fast on an 8 GB Mac, and it reads pictures too.
Limitations
It takes up to 30 seconds of audio at a time.
Video is limited to 60 seconds, looked at one frame per second.
It may state facts that are wrong or out of date.
Good to know
The E in E4B stands for effective.
Its training included web pages, code, images and audio, up to January 2025.
Sources: google/gemma-4-E4B-it model card, Outmake's speed estimates, LM Arena and Hugging Face. Every line comes from one of these.
Good at
Licence
Apache 2.0. Commercial use: yes. The maker's page has the full terms.
Gemma 4 E4B runs well from a Mac with an M1 and 8 GB of memory, at Q3_K_M, writing about 4.3 words (5.7 tokens) a second. 59 of 59 Mac configurations run it well.
Compare with
The nearest models in size, judged for the same Mac.
- Qwen3.5 9BChat that reads imagesRuns great · 6.6 GB · 13 words/sGet
- K2-Horizon 7BStrong at coding and multi-step work for a 16GB Mac, in words only.Runs great · 5.8 GB · 6.5 words/sGet
- Gemma 4 12BChat with pictures and audioRuns great · 5.9 GB · 10 words/sGet
- K2-Horizon 3.7BFor tasks that need tools called along the way, with reasoning worked through before each answer.Runs great · 4.4 GB · 7.4 words/sGet
Sizes
| Size | Download | Memory at 8K | Memory at 32K | Open |
|---|---|---|---|---|
| UD-Q5_K_XLBest | 7.6 GB | 5.3 GB | 5.5 GB | Open |
All 21 sizes
| Size | Download | Memory at 8K | Memory at 32K | Open |
|---|---|---|---|---|
| UD-IQ2_M | 4.5 GB | 4.5 GBestimated | 4.5 GB | Open |
| UD-IQ3_XXS | 4.7 GB | 4.7 GBestimated | 4.7 GB | Open |
| UD-Q2_K_XL | 4.7 GB | 4.7 GBestimated | 4.7 GB | Open |
| Q3_K_S | 4.9 GB | 4.9 GBestimated | 4.9 GB | Open |
| Q3_K_M | 5.0 GB | 5.0 GBestimated | 5.0 GB | Open |
| UD-Q3_K_XL | 5.6 GB | 5.6 GBestimated | 5.6 GB | Open |
| IQ4_XS | 5.7 GB | 5.7 GBestimated | 5.7 GB | Open |
| IQ4_NL | 5.8 GB | 5.8 GBestimated | 5.8 GB | Open |
| Q4_0 | 5.8 GB | 5.8 GBestimated | 5.8 GB | Open |
| Q4_K_S | 5.8 GB | 5.8 GBestimated | 5.8 GB | Open |
| Q4_K_M | 6.0 GB | 6.0 GBestimated | 6.0 GB | Open |
| Q4_1 | 6.1 GB | 6.1 GBestimated | 6.1 GB | Open |
| UD-Q4_K_XL | 6.1 GB | 6.1 GBestimated | 6.1 GB | Open |
| Q5_K_S | 6.4 GB | 6.4 GBestimated | 6.4 GB | Open |
| Q5_K_M | 6.5 GB | 6.5 GBestimated | 6.5 GB | Open |
| Q6_K | 8.1 GB | 8.1 GBestimated | 8.1 GB | Open |
| UD-Q6_K_XL | 8.4 GB | 8.4 GBestimated | 8.4 GB | Open |
| Q8_0 | 9.2 GB | 9.2 GBestimated | 9.2 GB | Open |
| UD-Q8_K_XL | 9.7 GB | 9.7 GBestimated | 9.7 GB | Open |
| BF16 | 16 GB | 16 GBestimated | 16 GB | Open |
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 E4B is an open model, so any app that runs GGUF files can run it. These take the UD-Q5_K_XL size from unsloth straight from Hugging Face.
- LM StudioMac, Windows, Linux
Open - JanMac, Windows, Linux
Open - MstyMac, Windows, Linux
Open - OllamaMac, Windows, Linux
ollama run hf.co/unsloth/gemma-4-E4B-it-GGUF:UD-Q5_K_XL - llama.cppMac, Windows, Linux
llama-server -hf unsloth/gemma-4-E4B-it-GGUF:UD-Q5_K_XL - Docker Model RunnerMac, Windows, Linux
docker model run hf.co/unsloth/gemma-4-E4B-it-GGUF:UD-Q5_K_XL
On a PC Graphics Card
| Video memory | Largest size that fits | Download | Memory at 8K |
|---|---|---|---|
| 6 GB | Q4_K_S | 5.8 GB | 5.8 GB |
| 8 GB | Q5_K_M | 6.5 GB | 6.5 GB |
| 10 GB | UD-Q8_K_XL | 9.7 GB | 9.7 GB |
| 11 GB | UD-Q8_K_XL | 9.7 GB | 9.7 GB |
| 12 GB | UD-Q8_K_XL | 9.7 GB | 9.7 GB |
| 16 GB | UD-Q8_K_XL | 9.7 GB | 9.7 GB |
| 20 GB | BF16 | 16 GB | 16 GB |
| 24 GB | BF16 | 16 GB | 16 GB |
| 32 GB | BF16 | 16 GB | 16 GB |
| 48 GB | BF16 | 16 GB | 16 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
| Mac | Size | Fit | Memory | Writes | Reads |
|---|---|---|---|---|---|
| M6 · 16 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 13 | 786 |
| M6 · 24 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 13 | 786 |
| M6 · 32 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 13 | 786 |
| M5 · 16 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 12 | 290 |
| M5 · 24 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 12 | 290 |
| M5 · 32 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 12 | 290 |
| M5 Pro · 24 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 23 | 580 |
| M5 Pro · 48 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 23 | 580 |
| M5 Pro · 64 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 23 | 726 |
| M5 Max · 36 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 33 | 1161 |
| M5 Max · 48 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 43 | 1451 |
| M5 Max · 64 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 43 | 1451 |
| M5 Max · 128 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 43 | 1451 |
| M5 Ultra · 96 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 77 | 2322 |
| M5 Ultra · 256 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 77 | 2902 |
| M5 Ultra · 512 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 77 | 2902 |
| M4 · 16 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 8.9 | 98 |
| M4 · 24 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 8.9 | 98 |
| M4 · 32 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 8.9 | 98 |
| M4 Pro · 24 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 18 | 195 |
| M4 Pro · 48 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 18 | 195 |
| M4 Pro · 64 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 18 | 195 |
| M4 Max · 36 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 25 | 390 |
| M4 Max · 48 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 31 | 488 |
| M4 Max · 64 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 31 | 488 |
| M4 Max · 96 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 25 | 390 |
| M4 Max · 128 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 31 | 488 |
| M3 · 8 GB | Q3_K_M | Runs well | 5.0 GB | 6.4 | 81 |
| M3 · 16 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 7.6 | 81 |
| M3 · 24 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 7.6 | 81 |
| M3 Pro · 18 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 11 | 141 |
| M3 Pro · 36 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 11 | 141 |
| M3 Max · 36 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 20 | 303 |
| M3 Max · 48 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 26 | 404 |
| M3 Max · 64 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 26 | 404 |
| M3 Max · 96 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 20 | 303 |
| M3 Max · 128 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 26 | 404 |
| M3 Ultra · 96 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 42 | 605 |
| M3 Ultra · 256 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 42 | 605 |
| M3 Ultra · 512 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 42 | 807 |
| M2 · 8 GB | Q3_K_M | Runs well | 5.0 GB | 6.4 | 79 |
| M2 · 16 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 7.6 | 79 |
| M2 · 24 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 7.6 | 79 |
| M2 Pro · 16 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 14 | 158 |
| M2 Pro · 32 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 14 | 158 |
| M2 Max · 32 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 26 | 297 |
| M2 Max · 64 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 26 | 297 |
| M2 Max · 96 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 26 | 376 |
| M2 Ultra · 64 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 42 | 594 |
| M2 Ultra · 128 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 42 | 594 |
| M2 Ultra · 192 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 42 | 594 |
| M1 · 8 GB | Q3_K_M | Runs well | 5.0 GB | 4.3 | 64 |
| M1 · 16 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 5.2 | 64 |
| M1 Pro · 16 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 13 | 128 |
| M1 Pro · 32 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 13 | 128 |
| M1 Max · 32 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 22 | 220 |
| M1 Max · 64 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 22 | 220 |
| M1 Ultra · 64 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 34 | 441 |
| M1 Ultra · 128 GB | UD-Q5_K_XL | Runs great | 7.6 GB | 34 | 441 |
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 E4B's Hugging Face card which Macs run it well, linked to this page. It updates every morning.
Sizes and files are read from unsloth/gemma-4-E4B-it-GGUF on Hugging Face and checked every morning. LM Arena has not rated this model, so the rating is Outmake's estimate from its size and release date; how the estimate is made. Catalog of November 4, 2026.