Llama on NVIDIA Ampere

Can I run Llama 4 Scout 109B-A17B on an RTX 3090?

Not at Q4_K_M — it needs 64.0 GB against 21.8 GB available. You would need 3 of these cards.

Won't fit

64.0 GB of 21.8 GB · 294%
067 GB
Weights 61.7 GB
KV cache 1.5 GB
Runtime overhead 0.8 GB
Over the limit 42.3 GB

Short by 42.3 GB. You can run it with 15 of 48 layers on the RTX 3090 and the rest in system RAM, at roughly 5.39 tok/s — usable for batch work, painful for chat. A smaller quantisation or a shorter context is usually the better trade.

Generation5.39tok/s
Prompt processing877tok/s
Max context0tokens
KV per 1K tokens0GB

Every quantisation of Llama 4 Scout 109B-A17B on a RTX 3090

Highlighted row is the highest quality that still fits at 8K context.

QuantisationbpwWeights TotalVerdictMax ctxtok/s
FP16 / BF16 16.00 203.0 GB 205.4 GB Won't fit — 1.33
INT8 / W8A8 8.50 107.4 GB 109.7 GB Won't fit — 2.69
Q8_0 (GGUF) 8.50 107.4 GB 109.7 GB Won't fit — 2.69
FP8 (E4M3) 8.00 101.5 GB 103.9 GB Won't fit — 2.91
Q6_K 6.56 83.2 GB 85.6 GB Won't fit — 3.68
Q5_K_M 5.67 71.9 GB 74.3 GB Won't fit — 4.33
Q5_K_S 5.52 70.0 GB 72.4 GB Won't fit — 4.55
Q4_K_M 4.85 61.7 GB 64.0 GB Won't fit — 5.39
Q4_K_S 4.58 58.3 GB 60.7 GB Won't fit — 5.68
Q4_0 4.55 58.0 GB 60.3 GB Won't fit — 5.87
AWQ 4-bit 4.25 56.8 GB 59.1 GB Won't fit — 5.98
GPTQ 4-bit 4.25 56.8 GB 59.1 GB Won't fit — 5.98
MXFP4 4.25 56.8 GB 59.1 GB Won't fit — 5.98
IQ4_XS 4.25 54.2 GB 56.6 GB Won't fit — 6.42
Q3_K_M 3.91 50.0 GB 52.3 GB Won't fit — 7.11
IQ3_M 3.70 47.4 GB 49.7 GB Won't fit — 7.69
IQ3_XXS 3.06 39.4 GB 41.8 GB Won't fit — 10.3
Q2_K 2.63 34.1 GB 36.4 GB Won't fit — 13.6
IQ2_XXS 2.06 27.0 GB 29.3 GB Won't fit — 22.9
IQ1_M 1.75 23.1 GB 25.4 GB Won't fit — 38.6

Also worth checking

Llama 4 Scout 109B-A17B on every GPU
Full hardware tableEverything that runs on a RTX 3090
Full model tableWhy the KV cache is the part that bites
Guide