Only with CPU offload
Llama 4 Scout 109B-A17B at Q4_K_M needs 63.4 GB but only 46.4 GB is addressable, so about 28% of the layers would stream from system RAM at roughly 60 GB/s. Expect around 4.5 tokens per second — usable for batch work, painful for chat.
A 10M-token window on paper, chunked attention in practice (8K chunks on 3 of 4 layers). Needs 64 GB+ at Q4.
Hardware requirements What hardware do I need to run Llama 4 Scout 109B-A17B? →At Q4_K_M with an 8K context: a 80 GB GPU (H100 SXM) or a 128 GB Mac. Minimum, recommended, high-quality and long-context tiers, with all 114 devices checked at every quantisation.
Fits instead: Llama 3.3 70B Instruct (42.8 GB) · Llama 3.1 8B Instruct (6.1 GB)
The VRAM budget
Quantisation ladder
| Quant | Weights | Total @ 8K | Max context | Tok/s | Quality | Fit |
|---|---|---|---|---|---|---|
| Q8_0 | 107.9 GB | 110.0 GB | — | ~1.3 | −0.1% ppl | 63.6 GB over |
| Q6_K | 83.2 GB | 85.3 GB | — | ~2.1 | −0.4% ppl | 38.9 GB over |
| Q5_K_M | 71.9 GB | 74.0 GB | — | ~2.9 | −0.8% ppl | 27.6 GB over |
| Q4_K_M | 61.3 GB | 63.4 GB | — | ~4.5 | −1.9% ppl | 17.0 GB over |
| Q3_K_M | 49.6 GB | 51.7 GB | — | ~11 | −5.4% ppl | 5.3 GB over |
| Q2_K | 42.5 GB | 44.6 GB | 46K | 30 | −15% ppl | Long context |
Quality is the published perplexity delta against f16 weights. Max context assumes an f16 KV cache; q8_0 roughly doubles it. This model interleaves sliding-window layers (8192 tokens, 12 of 48 layers global), which is why its cache barely grows with context.
How to run it
$ llama-server \
-hf meta-llama/Llama-4-Scout-17B-16E-Instruct:Q4_K_M \
-c 8192 -ngl 34
The engine underneath most of the others. Every knob is exposed. More on llama.cpp.
Related models on a Radeon Pro W7900
| Model | Total | Tok/s est. | |
|---|---|---|---|
| Llama 3.3 70B Instruct 70.6B · same family |
Q4_K_M · 42.8 GB | 13 | Tight fit |
| Llama 3.1 8B Instruct 8.03B · same family |
Q4_K_M · 6.1 GB | 116 | Runs great |
| Llama 3.2 3B Instruct 3.21B · same family |
Q4_K_M · 3.3 GB | 290 | Runs great |
| Llama 3.2 1B Instruct 1.24B · same family |
Q4_K_M · 1.5 GB | 750 | Runs great |
| gpt-oss 120B 117B MoE · similar size |
MXFP4 · 61.4 GB | ~18 | CPU offload |
| Mistral Small 4 119B-A6B 119B MoE · similar size |
Q4_K_M · 67.7 GB | ~10 | CPU offload |
| Nemotron 3 Super 120B-A12B 124B MoE · similar size |
Q4_K_M · 70.4 GB | ~5.3 | CPU offload |
Llama 4 Scout 109B-A17B on other hardware
| Device | Total | Tok/s est. | |
|---|---|---|---|
| GeForce RTX 4060 Laptop 7.0 GB usable · 256 GB/s |
63.4 GB | — | Won't fit |
| GeForce RTX 3060 12 GB 10.6 GB usable · 360 GB/s |
63.4 GB | — | Won't fit |
| M4 · 16 GB 10.7 GB usable · 120 GB/s |
63.4 GB | — | Won't fit |
| GeForce RTX 4090 22.4 GB usable · 1008 GB/s |
63.4 GB | — | Won't fit |
| Radeon RX 7900 XTX 22.4 GB usable · 960 GB/s |
63.4 GB | — | Won't fit |
| M3 Max · 64 GB 48.0 GB usable · 400 GB/s |
63.4 GB | — | Won't fit |
| H100 SXM 78.4 GB usable · 3350 GB/s |
63.4 GB | 82 | Runs great |
| DGX Spark · 128 GB 96.0 GB usable · 273 GB/s |
63.4 GB | 7.4 | Runs great |