Hugging FaceA 8B chat model from Sao10K.
- Downloads
- 78K
- Likes
- 120
- Parameters
- 8B
- Context
- 8K
- Licence
- cc-by-nc-4.0
Runs great on M4 · 16 GB
The Q6_K size from QuantFactory is the one to get: a 6.6 GB download that uses 7.3 GB of memory, and writes about 6.3 words a second.
L3 8B Stheno V3.2 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.
- Meta Llama 3.1 8Bcamenduru · 475K downloadsRuns great · 4.7 GB · 7.6 words/s
- Meta Llama 3 8BMeta · 134K downloadsRuns great · 6.6 GB · 6.3 words/s
- C4ai Command R7b 12 2024Cohere · 43K downloadsRuns great · 6.9 GB · 6.2 words/s
- Llama 3.1 8BMeta · 26K downloads · Arena 1211Runs great · 6.6 GB · 6.3 words/s
Sizes
From bartowski/L3-8B-Stheno-v3.2-GGUF on Hugging Face. A smaller size is a smaller download and a little less accurate.
| Size | Download | Memory at 8K | Get |
|---|---|---|---|
| IQ2_XS | 2.6 GB | 3.3 GBestimated | Open |
| IQ2_S | 2.8 GB | 3.5 GBestimated | Open |
| IQ2_M | 2.9 GB | 3.7 GBestimated | Open |
| IQ3_XXS | 3.3 GB | 4.0 GBestimated | Open |
| IQ3_XS | 3.5 GB | 4.2 GBestimated | Open |
| IQ3_M | 3.8 GB | 4.5 GBestimated | Open |
| IQ4_XS | 4.4 GB | 5.2 GBestimated | Open |
| F32 | 32 GB | 33 GBestimated | Open |
Run It Anywhere
L3 8B Stheno V3.2 is an open model, so any app that runs GGUF files can run it. These take the Q6_K size from QuantFactory straight from Hugging Face.
- LM StudioMac, Windows, Linux
Open - JanMac, Windows, Linux
Open - MstyMac, Windows, Linux
Open - OllamaMac, Windows, Linux
ollama run hf.co/QuantFactory/L3-8B-Stheno-v3.2-GGUF:Q6_K - llama.cppMac, Windows, Linux
llama-server -hf QuantFactory/L3-8B-Stheno-v3.2-GGUF:Q6_K - Docker Model RunnerMac, Windows, Linux
docker model run hf.co/QuantFactory/L3-8B-Stheno-v3.2-GGUF:Q6_K
On a PC Graphics Card
| Video memory | Largest size that fits | Download | Memory at 8K |
|---|---|---|---|
| 6 GB | IQ4_XS | 4.4 GB | 5.2 GB |
| 8 GB | IQ4_XS | 4.4 GB | 5.2 GB |
| 10 GB | IQ4_XS | 4.4 GB | 5.2 GB |
| 11 GB | IQ4_XS | 4.4 GB | 5.2 GB |
| 12 GB | IQ4_XS | 4.4 GB | 5.2 GB |
| 16 GB | IQ4_XS | 4.4 GB | 5.2 GB |
| 20 GB | IQ4_XS | 4.4 GB | 5.2 GB |
| 24 GB | IQ4_XS | 4.4 GB | 5.2 GB |
| 32 GB | IQ4_XS | 4.4 GB | 5.2 GB |
| 48 GB | F32 | 32 GB | 33 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 | Q6_K | Runs great | 7.3 GB | 9.2 |
| M6 · 24 GB | Q8_0 | Runs great | 9.3 GB | 8.9 |
| M6 · 32 GB | Q8_0 | Runs great | 9.3 GB | 8.9 |
| M5 · 16 GB | Q6_K | Runs great | 7.3 GB | 8.3 |
| M5 · 24 GB | Q8_0 | Runs great | 9.3 GB | 8.1 |
| M5 · 32 GB | Q8_0 | Runs great | 9.3 GB | 8.1 |
| M5 Pro · 24 GB | Q8_0 | Runs great | 9.3 GB | 16 |
| M5 Pro · 48 GB | Q8_0 | Runs great | 9.3 GB | 16 |
| M5 Pro · 64 GB | Q8_0 | Runs great | 9.3 GB | 16 |
| M5 Max · 36 GB | Q8_0 | Runs great | 9.3 GB | 24 |
| M5 Max · 48 GB | Q8_0 | Runs great | 9.3 GB | 31 |
| M5 Max · 64 GB | Q8_0 | Runs great | 9.3 GB | 31 |
| M5 Max · 128 GB | Q8_0 | Runs great | 9.3 GB | 31 |
| M5 Ultra · 96 GB | Q8_0 | Runs great | 9.3 GB | 59 |
| M5 Ultra · 256 GB | Q8_0 | Runs great | 9.3 GB | 59 |
| M5 Ultra · 512 GB | Q8_0 | Runs great | 9.3 GB | 59 |
| M4 · 16 GB | Q6_K | Runs great | 7.3 GB | 6.3 |
| M4 · 24 GB | Q8_0 | Runs great | 9.3 GB | 6.1 |
| M4 · 32 GB | Q8_0 | Runs great | 9.3 GB | 6.1 |
| M4 Pro · 24 GB | Q8_0 | Runs great | 9.3 GB | 13 |
| M4 Pro · 48 GB | Q8_0 | Runs great | 9.3 GB | 13 |
| M4 Pro · 64 GB | Q8_0 | Runs great | 9.3 GB | 13 |
| M4 Max · 36 GB | Q8_0 | Runs great | 9.3 GB | 19 |
| M4 Max · 48 GB | Q8_0 | Runs great | 9.3 GB | 24 |
| M4 Max · 64 GB | Q8_0 | Runs great | 9.3 GB | 24 |
| M4 Max · 96 GB | Q8_0 | Runs great | 9.3 GB | 19 |
| M4 Max · 128 GB | Q8_0 | Runs great | 9.3 GB | 24 |
| M3 · 8 GB | IQ4_XS | Runs well | 5.2 GB | 6.6 |
| M3 · 16 GB | Q6_K | Runs great | 7.3 GB | 5.3 |
| M3 · 24 GB | Q8_0 | Runs great | 9.3 GB | 5.2 |
| M3 Pro · 18 GB | Q6_K | Runs great | 7.3 GB | 7.9 |
| M3 Pro · 36 GB | Q8_0 | Runs great | 9.3 GB | 7.6 |
| M3 Max · 36 GB | Q8_0 | Runs great | 9.3 GB | 15 |
| M3 Max · 48 GB | Q8_0 | Runs great | 9.3 GB | 19 |
| M3 Max · 64 GB | Q8_0 | Runs great | 9.3 GB | 19 |
| M3 Max · 96 GB | Q8_0 | Runs great | 9.3 GB | 15 |
| M3 Max · 128 GB | Q8_0 | Runs great | 9.3 GB | 19 |
| M3 Ultra · 96 GB | Q8_0 | Runs great | 9.3 GB | 34 |
| M3 Ultra · 256 GB | Q8_0 | Runs great | 9.3 GB | 34 |
| M3 Ultra · 512 GB | Q8_0 | Runs great | 9.3 GB | 34 |
| M2 · 8 GB | IQ4_XS | Runs well | 5.2 GB | 6.6 |
| M2 · 16 GB | Q6_K | Runs great | 7.3 GB | 5.3 |
| M2 · 24 GB | Q8_0 | Runs great | 9.3 GB | 5.2 |
| M2 Pro · 16 GB | Q6_K | Runs great | 7.3 GB | 10 |
| M2 Pro · 32 GB | Q8_0 | Runs great | 9.3 GB | 10 |
| M2 Max · 32 GB | Q8_0 | Runs great | 9.3 GB | 19 |
| M2 Max · 64 GB | Q8_0 | Runs great | 9.3 GB | 19 |
| M2 Max · 96 GB | Q8_0 | Runs great | 9.3 GB | 19 |
| M2 Ultra · 64 GB | Q8_0 | Runs great | 9.3 GB | 34 |
| M2 Ultra · 128 GB | Q8_0 | Runs great | 9.3 GB | 34 |
| M2 Ultra · 192 GB | Q8_0 | Runs great | 9.3 GB | 34 |
| M1 · 8 GB | IQ4_XS | Runs well | 5.2 GB | 4.4 |
| M1 · 16 GB | Q6_K | Runs great | 7.3 GB | 3.6 |
| M1 Pro · 16 GB | Q6_K | Runs great | 7.3 GB | 9.9 |
| M1 Pro · 32 GB | Q8_0 | Runs great | 9.3 GB | 9.6 |
| M1 Max · 32 GB | Q8_0 | Runs great | 9.3 GB | 18 |
| M1 Max · 64 GB | Q8_0 | Runs great | 9.3 GB | 18 |
| M1 Ultra · 64 GB | Q8_0 | Runs great | 9.3 GB | 30 |
| M1 Ultra · 128 GB | Q8_0 | Runs great | 9.3 GB | 30 |
Writes is words a second, estimated from the Mac's memory speed and the model's size; how, and how wrong.
Packaged by
4 repositories on Hugging Face carry this model for llama.cpp. Outmake reads them all as one model.
Information
- Maker
- Sao10K
- Released
- June 2024
- Parameters
- 8B
- Longest chat
- 8K tokens
- Reads pictures
- No
- Licence
- cc-by-nc-4.0
- Downloads this month
- 77,797
- Likes
- 120
- Sizes
- 8, 2.6 to 32 GB
- Kind
- Maker's release
- Source
- Sao10K/L3-8B-Stheno-v3.2
Badge for the model card
Shows on L3 8B Stheno V3.2'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.