How much VRAM does Llama 3.3 70B need?

At Q4_K_M with 8K context: 43.2GB weights + 2.7GB KV cache + 1.5GB runtime overhead = 47.4GB total. RTX 4090 offers 24.0GB usable VRAM, so the verdict is: Not feasible.

What is the best quantization for Llama 3.3 70B on RTX 4090?

None — even Q2_K (32.2GB) exceeds this GPU's 24.0GB usable VRAM. Use a smaller model, multiple GPUs, or a cloud GPU.

Can RTX 4090 run Llama 3.3 70B?

❌ Not feasible

Llama 3.3 70B @ Q4_K_M · 8K context

Needed

47.4 GB

Usable

24.0 GB

VRAM breakdown

Weights (Q4_K_M)	43.2 GB
KV cache (8K context)	2.7 GB
Runtime overhead	1.5 GB
Total	47.4 GB

Computed at 8K context with fp16 KV cache. Longer contexts need more VRAM — use the VRAM calculator for other settings.

Recommended quantization

Does not fit at any quantization

2× RTX 4090 gives 48.0GB usable combined — enough to run it (Tight fit).

Smaller models this GPU runs well

Mistral Small 3.2 24B
gpt-oss-20b

FAQ

How much VRAM does Llama 3.3 70B need?: At Q4_K_M with 8K context: 43.2GB weights + 2.7GB KV cache + 1.5GB runtime overhead = 47.4GB total. RTX 4090 offers 24.0GB usable VRAM, so the verdict is: Not feasible.
What is the best quantization for Llama 3.3 70B on RTX 4090?: None — even Q2_K (32.2GB) exceeds this GPU's 24.0GB usable VRAM. Use a smaller model, multiple GPUs, or a cloud GPU.

Check another combination in the GPU Checker →

Data verified 2026-08-04