Can I Run Llama 3.3 on NVIDIA L40S?
Autor: Jakub Rusinowski · Ostatnia aktualizacja: 21 lipca 2026
Yes, comfortably — Llama 3.3 70B Instruct at Q2_K_XS (Tight) needs 23.7 GB at 8K context (20.2 GB weights + 3.5 GB KV cache/overhead), leaving ~24.3 GB of the NVIDIA L40S's 48 GB, at ~34 tok/s (est.).
See what else this hardware can run →
Jako partner Amazon zarabiamy na kwalifikujących się zakupach. Linki do chmurowych GPU są linkami polecającymi — możemy otrzymać prowizję bez dodatkowych kosztów dla Ciebie.
NVIDIA L40S Specs
| VRAM | 48 GB |
| Memory Bandwidth | 864 GB/s |
Llama 3.3 70B Instruct on the NVIDIA L40S: VRAM by quantization
| Quant | VRAM needed | Fits 48 GB? | Max context | Whole PC |
|---|---|---|---|---|
| F16 | 143.5 GB | ✗ No | — | — |
| Q8_0 | 77.9 GB | ✗ No | — | — |
| Q6_K | 60.9 GB | ✗ No | — | — |
| Q5_K_M | 53.1 GB | ✗ No | — | — |
| Q4_K_M | 45.7 GB | ✓ Yes | 8K | — |
| Q3_K_M | 33.3 GB | ✓ Yes | 32K | — |
| Q2_K | 26.5 GB | ✓ Yes | 64K | — |
Assumes an 8K-token context with an f16 KV cache. A longer window needs more; a quantized KV cache needs less. “Max context” is the largest window that still fits in 48 GB. Figures are estimates from parameter count, quantization and memory bandwidth — the analyzer lets you tune KV-cache quant and context.
Context costs VRAM too. At Q2_K_XS (Tight) on the NVIDIA L40S: 8K 23.7 GB · 32K 31.7 GB · 128K 64 GB ✗. Past 128K it no longer fits 48 GB — a q8_0 KV cache buys roughly half of that back, which the calculator will price for you.
No compatible GPU? Llama 3.3 70B Instruct on 256 GB of system RAM, CPU only: runs, but too slowly to be worth it.
Llama 3.3 Sizes That Fit the NVIDIA L40S (at 8K context)
| Llama 3.3 70B Instruct | Q2_K_XS (Tight) · 23.7 GB at 8K context · ~34 tok/s (est.) |
At 2 hrs/day, buying (~$7,499) beats renting at $0.34/hr after about 30.6 years.
Affiliate links — we may earn a commission if you sign up, at no extra cost to you.
Cloud rates verified 2026-07 — estimates, and marketplace prices vary. Buying price is GPU MSRP only, not a full PC.
FAQ
Will Llama 3.3 run on the NVIDIA L40S?
Yes, comfortably — Llama 3.3 70B Instruct at Q2_K_XS (Tight) needs 23.7 GB at 8K context (20.2 GB weights + 3.5 GB KV cache/overhead), leaving ~24.3 GB of the NVIDIA L40S's 48 GB, at ~34 tok/s (est.).
Which Llama 3.3 variant fits best on the NVIDIA L40S?
Llama 3.3 70B Instruct at Q2_K_XS (Tight) quantization needs 23.7 GB at 8K context (20.2 GB weights + 3.5 GB KV cache/overhead), estimated ~34 tokens/sec.
Llama 3.3 on Other GPUs
- Llama 3.3 on NVIDIA RTX 6000 Ada Generation
- Llama 3.3 on Apple M3 Pro
- Llama 3.3 on NVIDIA GeForce RTX 5090
- Llama 3.3 on Apple M4
- Llama 3.3 on Apple M2 Pro
Popular Models on the NVIDIA L40S
VRAM Tier
Troubleshooting
Buying Guide
← Can I Run It? | Llama 3.3 Model Page | NVIDIA L40S GPU Page | VRAM calculator | Check Your Hardware