Can I run Llama 4 Maverick 17B 128E Instruct locally?
Llama 4 Maverick 17B 128E Instruct by meta-llama has 402B parameters (18B active per token, 128 experts, 1 used per token). At 4-bit with an 8K context it needs about 228 GB of GPU memory. It fits on 3 of 65 GPUs and Macs we track.
Memory needed by quant and context length
GB of GPU memory = weights + KV cache (FP16) + runtime overhead. Weights use real file sizes from unsloth/Llama-4-Maverick-17B-128E-Instruct-GGUF where available (marked •), otherwise parameter count × bits per weight. How we calculate.
| Quant | Quality | 4K ctx | 8K ctx | 32K ctx | 128K ctx |
|---|---|---|---|---|---|
| 2-bit (Q2_K) • | rough | 137 GB | 138 GB | 142 GB | 160 GB |
| 3-bit (Q3_K_M) • | usable | 179 GB | 180 GB | 185 GB | 203 GB |
| 4-bit (Q4_K_M) • | good, most popular | 228 GB | 228 GB | 233 GB | 251 GB |
| 5-bit (Q5_K_M) • | very good | 266 GB | 267 GB | 272 GB | 290 GB |
| 6-bit (Q6_K) • | near lossless | 308 GB | 308 GB | 313 GB | 331 GB |
| 8-bit (Q8_0) • | effectively lossless | 398 GB | 399 GB | 403 GB | 421 GB |
| 16-bit (F16/BF16) • | original weights | 748 GB | 749 GB | 753 GB | 771 GB |
Which GPUs and Macs run Llama 4 Maverick 17B 128E Instruct?
Best quality quant that fits at 8K context. "Tight" = within 10% of usable memory. Speed is a ceiling from memory bandwidth (real speed is typically 50–80% of it).
Prefer an API?
Meta: Llama 4 Maverick is on OpenRouter at $0.35 per 1M input tokens and $1.00 per 1M output tokens. Price details and history →
Questions
How much VRAM does Llama 4 Maverick 17B 128E Instruct need?
About 228 GB at 4-bit (Q4_K_M) with an 8K context: 226 GB of weights (real GGUF file size), 1.5 GB of KV cache and 0.8 GB of runtime overhead. At 8-bit it needs about 399 GB.
What is the smallest GPU that runs Llama 4 Maverick 17B 128E Instruct?
No single consumer GPU in our table fits it at Q4_K_M with 8K context; the smallest option that does is the Mac M3 Ultra 512GB (512 GB).
Can I run Llama 4 Maverick 17B 128E Instruct on a Mac?
Yes. At Q4_K_M with 8K context it fits on a Mac M3 Ultra 512GB or larger, using the default macOS GPU memory limit.
Similar models
Same family or a similar parameter count, with the VRAM each needs at Q4_K_M and 8K context.
Llama 4 Scout 17B 16E Instruct
109B · 63.2 GB at 4-bitLlama 4 Scout 17B 16E
109B · 64.1 GB at 4-bitLlama 3.3 70B Instruct
71B · 43.3 GB at 4-bitMiniMax M3
427B · 248 GB at 4-bitGLM 4.5
358B · 205 GB at 4-bitGLM 4.7
358B · 205 GB at 4-bitModel facts
| Hugging Face repo | meta-llama/Llama-4-Maverick-17B-128E-Instruct |
| Parameters | 401,583,781,376 (402B) · ~18B active (computed from config) |
| Layers / hidden size | 48 / 5120 |
| Attention | GQA; 40 heads, 8 KV heads, head dim 128 |
| KV cache per token (FP16) | 192.0 KiB |
| Max context | 1,048,576 tokens |
| License | other (gated: accept terms on Hugging Face) |
| Released on HF | Apr 1, 2025 |
| Downloads (30d) / likes | 13,261 / 525 |
Sources: Hugging Face model API · config.json (unsloth/Llama-4-Maverick-17B-128E-Instruct-GGUF) · GGUF files · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.