Can I run Llama 4 Scout 17B 16E Instruct locally?
Llama 4 Scout 17B 16E Instruct by meta-llama has 109B parameters (18B active per token, 16 experts, 1 used per token). At 4-bit with an 8K context it needs about 63.2 GB of GPU memory. It fits on 22 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-Scout-17B-16E-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 | 38.4 GB | 39.1 GB | 43.6 GB | 61.6 GB |
| 3-bit (Q3_K_M) • | usable | 49.7 GB | 50.5 GB | 55.0 GB | 73.0 GB |
| 4-bit (Q4_K_M) • | good, most popular | 62.4 GB | 63.2 GB | 67.7 GB | 85.7 GB |
| 5-bit (Q5_K_M) • | very good | 72.8 GB | 73.6 GB | 78.1 GB | 96.1 GB |
| 6-bit (Q6_K) • | near lossless | 83.9 GB | 84.6 GB | 89.1 GB | 107 GB |
| 8-bit (Q8_0) • | effectively lossless | 108 GB | 109 GB | 113 GB | 131 GB |
| 16-bit (F16/BF16) • | original weights | 202 GB | 203 GB | 208 GB | 226 GB |
Which GPUs and Macs run Llama 4 Scout 17B 16E 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 Scout is on OpenRouter at $0.10 per 1M input tokens and $0.30 per 1M output tokens. Price details and history →
Questions
How much VRAM does Llama 4 Scout 17B 16E Instruct need?
About 63.2 GB at 4-bit (Q4_K_M) with an 8K context: 60.9 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 109 GB.
What is the smallest GPU that runs Llama 4 Scout 17B 16E Instruct?
No single consumer GPU in our table fits it at Q4_K_M with 8K context; the smallest option that does is the H100 PCIe (80 GB).
Can I run Llama 4 Scout 17B 16E Instruct on a Mac?
Yes. At Q4_K_M with 8K context it fits on a Mac M2 Max 96GB 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
109B · 64.1 GB at 4-bitLlama 3.3 70B Instruct
71B · 43.3 GB at 4-bitLlama 3.1 70B
71B · 42.7 GB at 4-bitGLM 4.5V
108B · 61.4 GB at 4-bitGLM 4.6V
108B · 67.8 GB at 4-bitGLM 4.5 Air
110B · 70.1 GB at 4-bitModel facts
| Hugging Face repo | meta-llama/Llama-4-Scout-17B-16E-Instruct |
| Parameters | 108,641,793,536 (109B) · ~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 | 10,485,760 tokens |
| License | other (gated: accept terms on Hugging Face) |
| Released on HF | Apr 2, 2025 |
| Downloads (30d) / likes | 110,352 / 1,391 |
Sources: Hugging Face model API · config.json (unsloth/Llama-4-Scout-17B-16E-Instruct-GGUF) · GGUF files · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.