Compatibility check
Can you run DeepSeek R1 Distill Llama 70B on the NVIDIA L40S?
Yes, with quantization. DeepSeek R1 Distill Llama 70B does not fit the NVIDIA L40S (48 GB) in FP16, but it runs at INT4 (4-bit) using about 41.9 GB (87% of VRAM). Use a GPTQ, AWQ, or GGUF build to get there, and keep prompts moderate to leave room for the KV cache.
Memory breakdown
Weights plus a 1.3 GB KV cache at 4,096tokens, against the card's 48 GB. Verdicts leave ~10% headroom for activations and fragmentation.
| Precision | Weights | KV cache | Total | % of 48 GB | Fit |
|---|---|---|---|---|---|
| FP16 / BF16full quality | 162.3 GB | 1.3 GB | 163.6 GB | 341% | No |
| INT8 (8-bit)near-full quality | 81.1 GB | 1.3 GB | 82.3 GB | 171% | No |
| INT4 (4-bit)GPTQ / AWQ / GGUF Q4 | 40.6 GB | 1.3 GB | 41.9 GB | 87% | Fits |
Planning estimates, not a substitute for profiling. Real usage varies with the inference runtime, batch size, and how much context you actually use — the KV cache grows linearly with prompt length.
GPUs that run DeepSeek R1 Distill Llama 70B
Cards where this model fits (at its best precision):
Models that fit the NVIDIA L40S
Other popular models that run on this card:
Go deeper
Frequently asked questions
Can the NVIDIA L40S run DeepSeek R1 Distill Llama 70B?
Not in FP16, but yes at INT4 (4-bit), where it uses about 41.9 GB versus the card's 48 GB.
How much VRAM does DeepSeek R1 Distill Llama 70B need?
Approximately 162.3 GB in FP16, 81.1 GB in INT8, and 40.6 GB in 4-bit for the weights, plus a KV cache of about 1.3 GB at 4,096 tokens.
Does quantization let DeepSeek R1 Distill Llama 70B fit on the NVIDIA L40S?
Yes. Dropping to INT4 (4-bit) brings total usage to about 41.9 GB, which fits the 48 GB card with headroom for the KV cache.
What happens to memory with longer context?
The KV cache grows linearly with prompt length. At 4,096 tokens it is about 1.3 GB here; doubling the context roughly doubles that term, so long-context use can push a tight fit over the edge.