unsloth
gemma 4 12b it
gemma 4 12b it is a 11.9B dense gemma4 model from unsloth. Configured context length is 262,144 tokens. License: apache-2.0. Quant sizes are the GGUF file bytes published on Hugging Face when a file exists.
Parameters11.9B
Context256K
Attentionhybrid
Licenseapache-2.0
48 layers · 8 KV heads · head dim 256 · 40 sliding layers · GGUF repo · config
Quant file sizes
| Quant | File size | Source |
|---|---|---|
| BF16 | 22.2 GB | Hugging Face file bytes |
| UD-Q8_K_XL | 12.7 GB | Hugging Face file bytes |
| Q8_0 | 11.8 GB | Hugging Face file bytes |
| UD-Q6_K_XL | 9.95 GB | Hugging Face file bytes |
| Q6_K | 9.11 GB | Hugging Face file bytes |
| UD-Q5_K_XL | 8.01 GB | Hugging Face file bytes |
| Q5_K_M | 7.84 GB | Hugging Face file bytes |
| Q5_K_S | 7.64 GB | Hugging Face file bytes |
| UD-Q4_K_XL | 6.86 GB | Hugging Face file bytes |
| Q4_K_M | 6.63 GB | Hugging Face file bytes |
| Q4_K_S | 6.30 GB | Hugging Face file bytes |
| IQ4_NL | 6.26 GB | Hugging Face file bytes |
| IQ4_XS | 5.94 GB | Hugging Face file bytes |
| Q4_1 | 6.89 GB | Hugging Face file bytes |
| Q4_0 | 6.28 GB | Hugging Face file bytes |
| UD-Q3_K_XL | 5.61 GB | Hugging Face file bytes |
| Q3_K_M | 5.30 GB | Hugging Face file bytes |
| Q3_K_S | 4.78 GB | Hugging Face file bytes |
| UD-IQ3_XXS | 4.32 GB | Hugging Face file bytes |
| UD-Q2_K_XL | 4.34 GB | Hugging Face file bytes |
On common machines
| Device | Verdict | Quant | Speed |
|---|---|---|---|
| GeForce RTX 4060 Ti 8GB | Tight fit | IQ4_XS | 20–27 |
| GeForce RTX 3060 12GB | Tight fit | Q6_K | 15–21 |
| GeForce RTX 4060 Ti 16GB | Tight fit | UD-Q8_K_XL | 9.4–13 |
| GeForce RTX 4090 | Runs | UD-Q8_K_XL | 33–45 |
| GeForce RTX 5090 | Runs | BF16 | 36–50 |
| Apple M4 Pro | Runs | UD-Q8_K_XL | 8.9–12 |
| Apple M5 | Tight fit | Q6_K | 6.5–8.8 |
| CPU only · DDR5-5600 | Runs on CPU | BF16 | 1.8–2.5 |
Smallest devices that fit it fully
- GeForce RTX 3050 6GB — Tight fit, UD-IQ3_XXS, 17–23 tok/s
- GeForce RTX 2060 6GB — Tight fit, UD-IQ3_XXS, 34–46 tok/s
- GeForce GTX 1660 Super — Tight fit, UD-IQ3_XXS, 34–46 tok/s
- GeForce GTX 1660 Ti — Tight fit, UD-IQ3_XXS, 29–40 tok/s
- Intel Arc A380 — Tight fit, UD-IQ3_XXS, 19–26 tok/s
- GeForce RTX 5060 Ti 8GB — Tight fit, IQ4_XS, 31–42 tok/s
Run it
Commands for the GeForce RTX 4060 Ti 16GB estimate (UD-Q8_K_XL).
llama.cpp
llama-server -hf unsloth/gemma-4-12b-it-GGUF:UD-Q8_K_XL -c 8192 -ngl 48
Ollama
ollama run gemma
LM Studio
lms get unsloth/gemma-4-12b-it-GGUF # Then load quant UD-Q8_K_XL