NVIDIA48 GB VRAM864 GB/s$0.7/hr

SmolLM2 1.7B Instruct on L40S 48GB: Does It Fit, How Fast, What It Costs

SmolLM2 1.7B Instruct at FP16 needs 3.4 GB for its weights and 16.3 GB in total at 8K context with 8 concurrent users. That means it fits on a single L40S 48GB. Every figure is computed by tokcalc's open-source capacity model.

Fits?
Yes, 1 GPU
16.3 GB needed / 48 GB available
Decode speed
165.2
tok/s per stream
Aggregate
1982
tok/s @ 8 concurrent
Cost / 1M tokens
$0.0981
1× L40S 48GB
What changes as you add L40S 48GBs
Tensor-parallel degree vs VRAM, throughput, and cost. Cost scales linearly with GPU count; decode throughput scales near-linearly too, so cost per token stays roughly flat.
GPUsTotal VRAMVRAM neededDecode tok/sAggregate tok/sTTFTCost / 1M tokFits?
1× L40S48 GB16.3 GB165.21982644.2 ms$0.0981✓
2× L40S96 GB16.3 GB280.86739337.6 ms$0.0577✓
4× L40S192 GB16.3 GB561.626957183.8 ms$0.0289✓
8× L40S384 GB16.3 GB1123107827106.9 ms$0.0144✓
Concurrency vs context length
On 1× L40S 48GB at FP16. KV cache scales linearly with context, so this is the number of users your GPUs can actually serve.
ContextConcurrent usersKV per requestFits?
4K55824.6 MB✓ fits
8K271.61 GB✓ fits
33K66.44 GB✓ fits
66K312.88 GB✓ fits
131K125.77 GB✓ fits
Other GPUs that run SmolLM2 1.7B Instruct
Same workload, same quantization — ranked by value (tokens per dollar-hour).
GPUGPUsAggregate tok/sCost / 1M tok
RTX 3090 24GB1×2147$0.0323
RTX 5090 32GB1×4111$0.0338
RTX 4090 24GB1×2312$0.0480
AMD MI300X 192GB1×12159$0.0571
V100 SXM2 32GB1×2065$0.0673
A100 40GB SXM41×3556$0.0703
Frequently asked questions

Can SmolLM2 1.7B Instruct run on L40S 48GB?

Yes. SmolLM2 1.7B Instruct at FP16 needs 16.3 GB including the KV cache, and L40S 48GB has 48 GB — so it fits on a single GPU.

How fast is SmolLM2 1.7B Instruct on L40S 48GB?

165.2 tokens/sec per stream and 1982 tok/s aggregate with continuous batching at 8 concurrent users. Time to first token is 644.2 ms, and inter-token latency is 6.1 ms.

What does SmolLM2 1.7B Instruct on L40S 48GB cost per million tokens?

$0.0981 per 1M output tokens at 1× L40S 48GB ($0.7000/hr). That is the marginal hardware cost — add idle capacity, ops, and egress for the real figure.

How much VRAM does SmolLM2 1.7B Instruct need on L40S 48GB?

3.4 GB for the weights at FP16, plus 12.9 GB of KV cache for 8 concurrent requests at 8K context — 16.3 GB total. Every extra concurrent user at long context costs more, so leave headroom.

Is L40S 48GB the right choice for SmolLM2 1.7B Instruct?

It works. Compare $0.0981 per 1M tokens against the alternatives below — the cheapest option that fits the same workload is 2× RTX 3080 10GB at $0.0187, and the fastest is 1× Cerebras CS-3 (WSE-3) at 45882 tok/s.

Adjust the workload yourself

Same formulas, your context length, concurrency, and quantisation.

Open in calculator