Hugging FaceA 71B chat model from Meta.
- Rating
- 3.2
- Downloads
- 175K
- Likes
- 290
- Parameters
- 71B
- Context
- 128K
- Licence
- llama3.3
Too big for this Mac
Its smallest size needs 41 GB of memory, more than M4 · 16 GB can give it.
Ratings
3.2 out of 5, across people's blind votes on its answers.
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.
Llama 3.3 70B runs well from a Mac with an M1 Pro and 32 GB of memory. 40 of 59 Mac configurations run it well.
Compare with
The nearest models in size.
- Meta Llama 3.1 70Bbartowski · 79K downloadsToo big for this Mac · 15 GB
- Llama 3.1 Nemotron 70BNVIDIA · 2.7K downloads · Arena 1299Too big for this Mac · 17 GB
- Llama 2 70BTheBloke · 18K downloads · Arena 1171Too big for this Mac · 15 GB
- Tulu 2 Dpo 70BAi2 · 936 downloads · Arena 1178Too big for this Mac · 29 GB
Sizes
From unsloth/Llama-3.3-70B-Instruct-GGUF on Hugging Face. A smaller size is a smaller download and a little less accurate.
| Size | Download | Memory at 8K | Get |
|---|---|---|---|
| UD-IQ1_S | 16 GB | 18 GBestimated | Open |
| UD-IQ1_M | 17 GB | 19 GBestimated | Open |
| UD-IQ2_XXS | 19 GB | 21 GBestimated | Open |
| UD-IQ2_M | 24 GB | 26 GBestimated | Open |
| Q2_K | 26 GB | 28 GBestimated | Open |
| Q2_K_L | 27 GB | 28 GBestimated | Open |
| UD-Q2_K_XL | 27 GB | 29 GBestimated | Open |
| UD-IQ3_XXS | 28 GB | 29 GBestimated | Open |
| Q3_K_S | 31 GB | 33 GBestimated | Open |
| Q3_K_M | 34 GB | 36 GBestimated | Open |
| UD-Q3_K_XL | 35 GB | 36 GBestimated | Open |
| IQ4_XS | 38 GB | 40 GBestimated | Open |
| IQ4_NL | 40 GB | 42 GBestimated | Open |
| Q4_0 | 40 GB | 42 GBestimated | Open |
| Q4_K_S | 40 GB | 42 GBestimated | Open |
| Q4_K_M | 43 GB | 44 GBestimated | Open |
| UD-Q4_K_XL | 43 GB | 44 GBestimated | Open |
| Q4_1 | 44 GB | 46 GBestimated | Open |
| Q5_K_S | 49 GB | 50 GBestimated | Open |
| UD-Q5_K_XL | 50 GB | 52 GBestimated | Open |
| Q5_K_M | 50 GB | 52 GBestimated | Open |
| Q6_K | 58 GB | 60 GBestimated | Open |
| UD-Q6_K_XL | 61 GB | 63 GBestimated | Open |
| Q8_0 | 75 GB | 77 GBestimated | Open |
| UD-Q8_K_XL | 81 GB | 83 GBestimated | Open |
| BF16 | 141 GB | 143 GBestimated | Open |
| F16 | 141 GB | 143 GBestimated | Open |
Run It Anywhere
Llama 3.3 70B is an open model, so any app that runs GGUF files can run it. These take the UD-IQ1_S 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/Llama-3.3-70B-Instruct-GGUF:UD-IQ1_S - llama.cppMac, Windows, Linux
llama-server -hf unsloth/Llama-3.3-70B-Instruct-GGUF:UD-IQ1_S - Docker Model RunnerMac, Windows, Linux
docker model run hf.co/unsloth/Llama-3.3-70B-Instruct-GGUF:UD-IQ1_S
On a PC Graphics Card
| Video memory | Largest size that fits | Download | Memory at 8K |
|---|---|---|---|
| 6 GB | None worth running | needs 26 GB | |
| 8 GB | None worth running | needs 26 GB | |
| 10 GB | None worth running | needs 26 GB | |
| 11 GB | None worth running | needs 26 GB | |
| 12 GB | None worth running | needs 26 GB | |
| 16 GB | None worth running | needs 26 GB | |
| 20 GB | None worth running | needs 26 GB | |
| 24 GB | None worth running | needs 26 GB | |
| 32 GB | UD-IQ3_XXS | 28 GB | 29 GB |
| 48 GB | Q4_1 | 44 GB | 46 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 |
|---|---|---|---|---|
| M6 · 16 GB | UD-IQ1_S | Too big | 41 GB | |
| M6 · 24 GB | IQ1_M | Tight fit | 18 GB | 3.5 |
| M6 · 32 GB | UD-IQ2_XXS | Runs well | 21 GB | 3.2 |
| M5 · 16 GB | UD-IQ1_S | Too big | 41 GB | |
| M5 · 24 GB | IQ1_M | Tight fit | 18 GB | 3.2 |
| M5 · 32 GB | UD-IQ2_XXS | Runs well | 21 GB | 2.9 |
| M5 Pro · 24 GB | IQ1_M | Tight fit | 18 GB | 6.3 |
| M5 Pro · 48 GB | UD-IQ3_XXS | Runs well | 31 GB | 4.5 |
| M5 Pro · 64 GB | Q4_K_S | Runs well | 42 GB | 3.4 |
| M5 Max · 36 GB | IQ2_S | Runs well | 24 GB | 7.8 |
| M5 Max · 48 GB | UD-IQ3_XXS | Runs well | 31 GB | 8.8 |
| M5 Max · 64 GB | Q4_K_S | Runs well | 42 GB | 6.7 |
| M5 Max · 128 GB | UD-Q6_K_XL | Runs great | 67 GB | 4.8 |
| M5 Ultra · 96 GB | Q4_1 | Runs great | 46 GB | 14 |
| M5 Ultra · 256 GB | Q8_0 | Runs great | 100 GB | 9.4 |
| M5 Ultra · 512 GB | Q8_0 | Runs great | 100 GB | 9.4 |
| M4 · 16 GB | UD-IQ1_S | Too big | 41 GB | |
| M4 · 24 GB | IQ1_M | Tight fit | 18 GB | 2.5 |
| M4 · 32 GB | UD-IQ2_XXS | Runs well | 21 GB | 2.3 |
| M4 Pro · 24 GB | IQ1_M | Tight fit | 18 GB | 5.3 |
| M4 Pro · 48 GB | UD-IQ3_XXS | Runs well | 31 GB | 3.9 |
| M4 Pro · 64 GB | Q4_K_S | Runs well | 42 GB | 3 |
| M4 Max · 36 GB | IQ2_S | Runs well | 24 GB | 6.5 |
| M4 Max · 48 GB | UD-IQ3_XXS | Runs well | 31 GB | 7.3 |
| M4 Max · 64 GB | Q4_K_S | Runs well | 42 GB | 5.7 |
| M4 Max · 96 GB | Q4_1 | Runs great | 46 GB | 4.7 |
| M4 Max · 128 GB | UD-Q6_K_XL | Runs great | 67 GB | 4.2 |
| M3 · 8 GB | UD-IQ1_S | Too big | 41 GB | |
| M3 · 16 GB | UD-IQ1_S | Too big | 41 GB | |
| M3 · 24 GB | IQ1_M | Tight fit | 18 GB | 2.1 |
| M3 Pro · 18 GB | UD-IQ1_S | Too big | 41 GB | |
| M3 Pro · 36 GB | IQ2_S | Runs well | 24 GB | 2.6 |
| M3 Max · 36 GB | IQ2_S | Runs well | 24 GB | 5 |
| M3 Max · 48 GB | UD-IQ3_XXS | Runs well | 31 GB | 5.7 |
| M3 Max · 64 GB | Q4_K_S | Runs well | 42 GB | 4.3 |
| M3 Max · 96 GB | Q4_1 | Runs great | 46 GB | 3.5 |
| M3 Max · 128 GB | UD-Q6_K_XL | Runs great | 67 GB | 3.1 |
| M3 Ultra · 96 GB | Q4_1 | Runs great | 46 GB | 8.6 |
| M3 Ultra · 256 GB | Q8_0 | Runs great | 100 GB | 5.8 |
| M3 Ultra · 512 GB | Q8_0 | Runs great | 100 GB | 5.8 |
| M2 · 8 GB | UD-IQ1_S | Too big | 41 GB | |
| M2 · 16 GB | UD-IQ1_S | Too big | 41 GB | |
| M2 · 24 GB | IQ1_M | Tight fit | 18 GB | 2.1 |
| M2 Pro · 16 GB | UD-IQ1_S | Too big | 41 GB | |
| M2 Pro · 32 GB | UD-IQ2_XXS | Runs well | 21 GB | 3.7 |
| M2 Max · 32 GB | UD-IQ2_XXS | Runs well | 21 GB | 7 |
| M2 Max · 64 GB | Q4_K_S | Runs well | 42 GB | 4.3 |
| M2 Max · 96 GB | Q4_1 | Runs great | 46 GB | 4.6 |
| M2 Ultra · 64 GB | Q4_K_S | Runs well | 42 GB | 8.1 |
| M2 Ultra · 128 GB | UD-Q6_K_XL | Runs great | 67 GB | 6 |
| M2 Ultra · 192 GB | Q8_0 | Runs great | 100 GB | 5.8 |
| M1 · 8 GB | UD-IQ1_S | Too big | 41 GB | |
| M1 · 16 GB | UD-IQ1_S | Too big | 41 GB | |
| M1 Pro · 16 GB | UD-IQ1_S | Too big | 41 GB | |
| M1 Pro · 32 GB | UD-IQ2_XXS | Runs well | 21 GB | 3.6 |
| M1 Max · 32 GB | UD-IQ2_XXS | Runs well | 21 GB | 6.5 |
| M1 Max · 64 GB | Q4_K_S | Runs well | 42 GB | 4.1 |
| M1 Ultra · 64 GB | Q4_K_S | Runs well | 42 GB | 7.5 |
| M1 Ultra · 128 GB | UD-Q6_K_XL | Runs great | 67 GB | 5.6 |
Writes is words a second, estimated from the Mac's memory speed and the model's size; how, and how wrong.
Packaged by
5 repositories on Hugging Face carry this model for llama.cpp. Outmake reads them all as one model.
Information
- Maker
- Meta
- Released
- November 2024
- Parameters
- 71B
- Longest chat
- 128K tokens
- Reads pictures
- No
- Licence
- llama3.3
- Downloads this month
- 174,843
- Likes
- 290
- Sizes
- 27, 16 to 141 GB
- Kind
- Maker's release
Badge for the model card
Shows on Llama 3.3 70B's Hugging Face card which Macs run it well, linked to this page. It updates every morning.
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.