NVIDIA32 GB VRAM1,792 GB/s$0.5/hr

OLMo 2 13B on RTX 5090 32GB: 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 needs 2× RTX 5090 32GB. Every figure is computed by tokcalc's open-source capacity model.

Fits?
Yes, 2×
36.7 GB needed / 64 GB available
Decode speed
76.16
tok/s per stream
Aggregate
1828
tok/s @ 8 concurrent
Cost / 1M tokens
$0.1520
2× RTX 5090 32GB

RTX 5090 32GB has no NVLink, so 2-way tensor parallelism runs over PCIe. tokcalc applies a flat 0.85× multi-GPU efficiency factor for this, which is realistic for NVLink-equipped datacenter cards but optimistic for PCIe scale-out — treat the throughput and cost figures for this configuration as an upper bound, and measure before committing.

What changes as you add RTX 5090 32GBs
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× RTX32 GB36.7 GB0.0000.000——✗
2× RTX64 GB36.7 GB76.1618282.07 s$0.1520✓
4× RTX128 GB36.7 GB152.373111.05 s$0.0760✓
8× RTX256 GB36.7 GB304.629245540.4 ms$0.0380✓
Concurrency vs context length
On 2× RTX 5090 32GB at FP16. KV cache scales linearly with context, so this is the number of users your GPUs can actually serve.
ContextConcurrent usersKV per requestFits?
4K56687.2 MB✓ fits
8K281.34 GB✓ fits
33K75.37 GB✓ fits
66K310.74 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
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
H100 SXM5 80GB1×1005$0.6910
Frequently asked questions

Can OLMo 2 13B run on RTX 5090 32GB?

Yes, but it needs 2× RTX 5090 32GB. At FP16 the total footprint is 36.7 GB (26.0 GB of weights plus 10.7 GB of KV cache at 8K × 8 concurrent users), against 64 GB of available VRAM across 2 GPUs.

How fast is OLMo 2 13B on RTX 5090 32GB?

76.16 tokens/sec per stream and 1828 tok/s aggregate with continuous batching at 8 concurrent users. Time to first token is 2.07 s, and inter-token latency is 13.1 ms.

What does OLMo 2 13B on RTX 5090 32GB cost per million tokens?

$0.1520 per 1M output tokens at 2× RTX 5090 32GB ($0.5000/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 RTX 5090 32GB?

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 RTX 5090 32GB the right choice for OLMo 2 13B?

It works. Compare $0.1520 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