unsloth
Llama 4 Maverick 17B 128E Instruct
Llama 4 Maverick 17B 128E Instruct is a 400.7B mixture-of-experts llama4 model from unsloth, with about 63.2B parameters active per token. Configured context length is 1,048,576 tokens. License: other. Quant sizes are the GGUF file bytes published on Hugging Face when a file exists.
Parameters401B / 63.2B active
Context1024K
Attentiongqa
Licenseother
48 layers · 8 KV heads · head dim 128 · GGUF repo · config
Quant file sizes
| Quant | File size | Source |
|---|---|---|
| BF16 | 746.4 GB | Hugging Face file bytes |
| UD-Q8_K_XL | 428.4 GB | Hugging Face file bytes |
| Q8_0 | 396.6 GB | Hugging Face file bytes |
| UD-Q6_K_XL | 317.7 GB | Hugging Face file bytes |
| Q6_K | 306.2 GB | Hugging Face file bytes |
| UD-Q5_K_XL | 267.3 GB | Hugging Face file bytes |
| Q5_K_M | 264.9 GB | Hugging Face file bytes |
| Q5_K_S | 256.8 GB | Hugging Face file bytes |
| UD-Q4_K_XL | 216.2 GB | Hugging Face file bytes |
| Q4_K_M | 226.1 GB | Hugging Face file bytes |
| Q4_K_S | 212.2 GB | Hugging Face file bytes |
| IQ4_NL | 210.3 GB | Hugging Face file bytes |
| IQ4_XS | 199.6 GB | Hugging Face file bytes |
| Q4_1 | 233.5 GB | Hugging Face file bytes |
| Q4_0 | 211.2 GB | Hugging Face file bytes |
| UD-Q3_K_XL | 167.2 GB | Hugging Face file bytes |
| Q3_K_M | 177.9 GB | Hugging Face file bytes |
| Q3_K_S | 160.8 GB | Hugging Face file bytes |
| UD-IQ3_XXS | 157.7 GB | Hugging Face file bytes |
| UD-Q2_K_XL | 142.2 GB | Hugging Face file bytes |
| Q2_K | 271.5 GB | Hugging Face file bytes |
| UD-IQ2_XXS | 125.9 GB | Hugging Face file bytes |
| UD-IQ1_M | 118.8 GB | Hugging Face file bytes |
| UD-IQ1_S | 112.5 GB | Hugging Face file bytes |
On common machines
| Device | Verdict | Quant | Speed |
|---|---|---|---|
| GeForce RTX 4060 Ti 8GB | Won't fit | UD-IQ1_S | — |
| GeForce RTX 3060 12GB | Won't fit | UD-IQ1_S | — |
| GeForce RTX 4060 Ti 16GB | Won't fit | UD-IQ1_S | — |
| GeForce RTX 4090 | Won't fit | UD-IQ1_S | — |
| GeForce RTX 5090 | Won't fit | UD-IQ1_S | — |
| Apple M4 Pro | Won't fit | UD-IQ1_S | — |
| Apple M5 | Won't fit | UD-IQ1_S | — |
| CPU only · DDR5-5600 | Won't fit | UD-IQ1_S | — |
Smallest devices that fit it fully
- NVIDIA H200 SXM — Tight fit, UD-IQ2_XXS, 115–157 tok/s
- NVIDIA B200 — Tight fit, UD-Q3_K_XL, 147–200 tok/s
Run it
Commands for the GeForce RTX 4060 Ti 16GB estimate (UD-IQ1_S).
llama.cpp
llama-server -hf unsloth/Llama-4-Maverick-17B-128E-Instruct-GGUF:UD-IQ1_S -c 2048 -ngl 0
No matching name in the Ollama library snapshot.
LM Studio
lms get unsloth/Llama-4-Maverick-17B-128E-Instruct-GGUF # Then load quant UD-IQ1_S