Can I Run Cogito v1 3B on NVIDIA L40S?
Written by Jakub Rusinowski · Last updated July 21, 2026
Yes. Cogito v1 3B needs 3.6 GB at 8K context (1.8 GB weights + 1.8 GB KV cache/overhead) at Q4_K_M and the NVIDIA L40S has 48 GB, leaving about 44.4 GB of headroom, at roughly 196 tok/s (estimated) with room for up to 128K of context.
See what else this hardware can run →
The numbers
| Model | Cogito v1 3B |
| Parameters | 3B |
| Quantization | Q4_K_M |
| VRAM needed (8K context) | 3.6 GB |
| NVIDIA L40S VRAM | 48 GB |
| Verdict | Runs |
| Max context | 128K |
| Estimated speed | ~196 tok/s |
Cogito v1 3B on the NVIDIA L40S: VRAM by quantization
| Quant | VRAM needed | Fits 48 GB? | Max context | Whole PC |
|---|---|---|---|---|
| F16 | 7.7 GB | ✓ Yes | 128K | — |
| Q8_0 | 4.9 GB | ✓ Yes | 128K | — |
| Q6_K | 4.2 GB | ✓ Yes | 128K | — |
| Q5_K_M | 3.9 GB | ✓ Yes | 128K | — |
| Q4_K_M | 3.6 GB | ✓ Yes | 128K | — |
| Q3_K_M | 3 GB | ✓ Yes | 128K | — |
| Q2_K | 2.7 GB | ✓ Yes | 128K | — |
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 Q4_K_M on the NVIDIA L40S: 8K 3.6 GB · 32K 6.4 GB · 128K 17.6 GB. Every window this model publishes still fits 48 GB.
What to do next
Nothing to change — run it at Q4_K_M. If you want more capability, check whether the next size up also fits.
No compatible GPU? Cogito v1 3B on 8 GB of system RAM, CPU only: runs.
Running Cogito v1 3B on this card:
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.
Other Cogito v1 sizes on the NVIDIA L40S
- Cogito v1 70B — runs (45.7 GB)
- Cogito v1 32B — runs (22.3 GB)
- Cogito v1 14B — runs (10.9 GB)
- Cogito v1 8B — runs (6.7 GB)
Cogito v1 3B on nearby hardware
- Apple M3 Pro (36 GB) — runs
- Apple M1 Max (64 GB) — runs
All Cogito v1 sizes on the NVIDIA L40S | Cogito v1 3B model page | NVIDIA L40S GPU page | VRAM calculator | Check your hardware