Cerebras44 GB VRAM20,000 GB/s

OLMo 2 13B on Cerebras CS-3 (WSE-3): 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 Cerebras CS-3 (WSE-3). Every figure is computed by tokcalc's open-source capacity model.

Fits?
Yes, 1 GPU
36.7 GB needed / 44 GB available
Decode speed
500.0
tok/s per stream
Aggregate
6000
tok/s @ 8 concurrent
Cost / 1M tokens
—
1× Cerebras CS-3 (WSE-3) · quote required
What changes as you add Cerebras CS-3 (WSE-3)s
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× Cerebras44 GB36.7 GB500.06000372.8 ms—✓
2× Cerebras88 GB36.7 GB850.020400201.6 ms—✓
4× Cerebras176 GB36.7 GB170081600115.8 ms—✓
8× Cerebras352 GB36.7 GB340032640072.9 ms—✓
Concurrency vs context length
On 1× Cerebras CS-3 (WSE-3) at FP16. KV cache scales linearly with context, so this is the number of users your GPUs can actually serve.
ContextConcurrent usersKV per requestFits?
4K26687.2 MB✓ fits
8K131.34 GB✓ fits
33K35.37 GB✓ fits
66K110.74 GB✓ fits
131K021.47 GB✗ does not fit
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 Cerebras CS-3 (WSE-3)?

Yes. OLMo 2 13B at FP16 needs 36.7 GB including the KV cache, and Cerebras CS-3 (WSE-3) has 44 GB — so it fits on a single GPU.

How fast is OLMo 2 13B on Cerebras CS-3 (WSE-3)?

500.0 tokens/sec per stream and 6000 tok/s aggregate with continuous batching at 8 concurrent users. Time to first token is 372.8 ms, and inter-token latency is 2.0 ms.

What does OLMo 2 13B on Cerebras CS-3 (WSE-3) cost per million tokens?

— per 1M output tokens at 1× Cerebras CS-3 (WSE-3). 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 Cerebras CS-3 (WSE-3)?

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 Cerebras CS-3 (WSE-3) the right choice for OLMo 2 13B?

It works. Compare — 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