NVIDIA48 GB VRAM864 GB/s$0.7/hr

OLMo 2 13B on L40S 48GB: Does It Fit, How Fast, What It Costs

OLMo 2 13B at FP16 needs 26.0 GB for its weights and 36.7 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
36.7 GB needed / 48 GB available
Decode speed
21.60
tok/s per stream
Aggregate
259.2
tok/s @ 8 concurrent
Cost / 1M tokens
$0.7502
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 GB36.7 GB21.60259.24.73 s$0.7502✓
2× L40S96 GB36.7 GB36.72881.32.38 s$0.4413✓
4× L40S192 GB36.7 GB73.4435251.21 s$0.2206✓
8× L40S384 GB36.7 GB146.914100618.1 ms$0.1103✓
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?
4K32687.2 MB✓ fits
8K161.34 GB✓ fits
33K45.37 GB✓ fits
66K210.74 GB✓ fits
131K121.47 GB✓ fits
Other GPUs that run OLMo 2 13B
Same workload, same quantization — ranked by value (tokens per dollar-hour).
GPUGPUsAggregate tok/sCost / 1M tok
RTX 5090 32GB2×1828$0.1520
AMD MI300X 192GB1×1590$0.4368
A100 40GB SXM41×465.0$0.5376
AMD MI325X 288GB1×1800$0.5401
A100 80GB SXM41×600.0$0.5556
Intel Gaudi 3 128GB1×990.0$0.5612
Frequently asked questions

Can OLMo 2 13B run on L40S 48GB?

Yes. OLMo 2 13B at FP16 needs 36.7 GB including the KV cache, and L40S 48GB has 48 GB — so it fits on a single GPU.

How fast is OLMo 2 13B on L40S 48GB?

21.60 tokens/sec per stream and 259.2 tok/s aggregate with continuous batching at 8 concurrent users. Time to first token is 4.73 s, and inter-token latency is 46.3 ms.

What does OLMo 2 13B on L40S 48GB cost per million tokens?

$0.7502 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 OLMo 2 13B need on L40S 48GB?

26.0 GB for the weights at FP16, plus 10.7 GB of KV cache for 8 concurrent requests at 8K context — 36.7 GB total. Every extra concurrent user at long context costs more, so leave headroom.

Is L40S 48GB the right choice for OLMo 2 13B?

It works. Compare $0.7502 per 1M tokens against the alternatives below — the cheapest option that fits the same workload is 4× RTX 3080 10GB at $0.0717, and the fastest is 1× Cerebras CS-3 (WSE-3) at 6000 tok/s.

Adjust the workload yourself

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

Open in calculator