Llama on NVIDIA Ampere

Can I run Llama 4 Scout 109B-A17B on an RTX 3080 10GB?

Not at Q4_K_M — it needs 64.0 GB against 8.6 GB available. You would need 8 of these cards.

Won't fit

64.0 GB of 8.6 GB · 745%
067 GB
Weights 61.7 GB
KV cache 1.5 GB
Runtime overhead 0.8 GB
Over the limit 55.4 GB

Short by 55.4 GB. You can run it with 4 of 48 layers on the RTX 3080 10GB and the rest in system RAM, at roughly 4.16 tok/s — usable for batch work, painful for chat. A smaller quantisation or a shorter context is usually the better trade.

Generation4.16tok/s
Prompt processing741tok/s
Max context0tokens
KV per 1K tokens0GB

Every quantisation of Llama 4 Scout 109B-A17B on a RTX 3080 10GB

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 — 1.26
INT8 / W8A8 8.50 107.4 GB 109.7 GB Won't fit — 2.37
Q8_0 (GGUF) 8.50 107.4 GB 109.7 GB Won't fit — 2.37
FP8 (E4M3) 8.00 101.5 GB 103.9 GB Won't fit — 2.50
Q6_K 6.56 83.2 GB 85.6 GB Won't fit — 3.08
Q5_K_M 5.67 71.9 GB 74.3 GB Won't fit — 3.60
Q5_K_S 5.52 70.0 GB 72.4 GB Won't fit — 3.69
Q4_K_M 4.85 61.7 GB 64.0 GB Won't fit — 4.16
Q4_K_S 4.58 58.3 GB 60.7 GB Won't fit — 4.47
Q4_0 4.55 58.0 GB 60.3 GB Won't fit — 4.49
AWQ 4-bit 4.25 56.8 GB 59.1 GB Won't fit — 4.58
GPTQ 4-bit 4.25 56.8 GB 59.1 GB Won't fit — 4.58
MXFP4 4.25 56.8 GB 59.1 GB Won't fit — 4.58
IQ4_XS 4.25 54.2 GB 56.6 GB Won't fit — 4.78
Q3_K_M 3.91 50.0 GB 52.3 GB Won't fit — 5.25
IQ3_M 3.70 47.4 GB 49.7 GB Won't fit — 5.52
IQ3_XXS 3.06 39.4 GB 41.8 GB Won't fit — 6.65
Q2_K 2.63 34.1 GB 36.4 GB Won't fit — 7.73
IQ2_XXS 2.06 27.0 GB 29.3 GB Won't fit — 10.1
IQ1_M 1.75 23.1 GB 25.4 GB Won't fit — 12.1

Also worth checking

Llama 4 Scout 109B-A17B on every GPU
Full hardware tableEverything that runs on a RTX 3080 10GB
Full model tableWhy the KV cache is the part that bites
Guide