unsloth
granite 4.0 h micro
granite 4.0 h micro is a 3.19B dense granitehybrid model from unsloth. Configured context length is 1,048,576 tokens. License: apache-2.0. Quant sizes are the GGUF file bytes published on Hugging Face when a file exists.
Parameters3.2B
Context1024K
Attentionhybrid
Licenseapache-2.0
40 layers · 8 KV heads · head dim 64 · 36 linear-attention layers · GGUF repo · config
Quant file sizes
| Quant | File size | Source |
|---|---|---|
| BF16 | 5.95 GB | Hugging Face file bytes |
| UD-Q8_K_XL | 3.56 GB | Hugging Face file bytes |
| Q8_0 | 3.16 GB | Hugging Face file bytes |
| UD-Q6_K_XL | 2.57 GB | Hugging Face file bytes |
| Q6_K | 2.44 GB | Hugging Face file bytes |
| UD-Q5_K_XL | 2.12 GB | Hugging Face file bytes |
| Q5_K_M | 2.12 GB | Hugging Face file bytes |
| Q5_K_S | 2.08 GB | Hugging Face file bytes |
| UD-Q4_K_XL | 1.83 GB | Hugging Face file bytes |
| Q4_K_M | 1.81 GB | Hugging Face file bytes |
| Q4_K_S | 1.74 GB | Hugging Face file bytes |
| IQ4_NL | 1.73 GB | Hugging Face file bytes |
| IQ4_XS | 1.64 GB | Hugging Face file bytes |
| Q4_1 | 1.90 GB | Hugging Face file bytes |
| Q4_0 | 1.73 GB | Hugging Face file bytes |
| UD-Q3_K_XL | 1.47 GB | Hugging Face file bytes |
| Q3_K_M | 1.45 GB | Hugging Face file bytes |
| Q3_K_S | 1.36 GB | Hugging Face file bytes |
| UD-IQ3_XXS | 1.21 GB | Hugging Face file bytes |
| UD-Q2_K_XL | 1.16 GB | Hugging Face file bytes |
| Q2_K | 2.28 GB | Hugging Face file bytes |
| UD-IQ2_XXS | 0.88 GB | Hugging Face file bytes |
| UD-IQ1_M | 0.78 GB | Hugging Face file bytes |
| UD-IQ1_S | 0.72 GB | Hugging Face file bytes |
On common machines
| Device | Verdict | Quant | Speed |
|---|---|---|---|
| GeForce RTX 4060 Ti 8GB | Tight fit | BF16 | 25–33 |
| GeForce RTX 3060 12GB | Runs great | BF16 | 31–42 |
| GeForce RTX 4060 Ti 16GB | Runs great | BF16 | 25–33 |
| GeForce RTX 4090 | Runs great | BF16 | 86–117 |
| GeForce RTX 5090 | Runs great | BF16 | 153–208 |
| Apple M4 Pro | Runs great | BF16 | 23–32 |
| Apple M5 | Runs great | BF16 | 13–18 |
| CPU only · DDR5-5600 | Runs on CPU | BF16 | 7.6–10 |
Smallest devices that fit it fully
- GeForce RTX 3050 6GB — Runs, UD-Q8_K_XL, 24–32 tok/s
- GeForce RTX 2060 6GB — Runs, UD-Q8_K_XL, 47–65 tok/s
- GeForce GTX 1660 Super — Runs, UD-Q8_K_XL, 47–65 tok/s
- GeForce GTX 1660 Ti — Runs, UD-Q8_K_XL, 41–55 tok/s
- Intel Arc A380 — Runs, UD-Q8_K_XL, 26–36 tok/s
- GeForce RTX 5060 Ti 8GB — Tight fit, BF16, 38–52 tok/s
Run it
Commands for the GeForce RTX 4060 Ti 16GB estimate (BF16).
llama.cpp
llama-server -hf unsloth/granite-4.0-h-micro-GGUF:BF16 -c 8192 -ngl 40
No matching name in the Ollama library snapshot.
LM Studio
lms get unsloth/granite-4.0-h-micro-GGUF # Then load quant BF16