Google95 GB VRAM819 GB/s$4.2/hr

OLMo 2 13B on TPU v5p (per-chip, 95GB): 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 TPU v5p (per-chip, 95GB). Every figure is computed by tokcalc's open-source capacity model.

Fits?
Yes, 1 GPU
36.7 GB needed / 95 GB available
Decode speed
20.48
tok/s per stream
Aggregate
245.7
tok/s @ 8 concurrent
Cost / 1M tokens
$4.75
1× TPU v5p (per-chip, 95GB)
What changes as you add TPU v5p (per-chip, 95GB)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× TPU95 GB36.7 GB20.48245.71.01 s$4.75✓
2× TPU190 GB36.7 GB34.81835.4522.8 ms$2.79✓
4× TPU380 GB36.7 GB69.613342276.4 ms$1.40✓
8× TPU760 GB36.7 GB139.213366153.2 ms$0.6983✓
Concurrency vs context length
On 1× TPU v5p (per-chip, 95GB) at FP16. KV cache scales linearly with context, so this is the number of users your GPUs can actually serve.
ContextConcurrent usersKV per requestFits?
4K102687.2 MB✓ fits
8K511.34 GB✓ fits
33K125.37 GB✓ fits
66K610.74 GB✓ fits
131K321.47 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 TPU v5p (per-chip, 95GB)?

Yes. OLMo 2 13B at FP16 needs 36.7 GB including the KV cache, and TPU v5p (per-chip, 95GB) has 95 GB — so it fits on a single GPU.

How fast is OLMo 2 13B on TPU v5p (per-chip, 95GB)?

20.48 tokens/sec per stream and 245.7 tok/s aggregate with continuous batching at 8 concurrent users. Time to first token is 1.01 s, and inter-token latency is 48.8 ms.

What does OLMo 2 13B on TPU v5p (per-chip, 95GB) cost per million tokens?

$4.75 per 1M output tokens at 1× TPU v5p (per-chip, 95GB) ($4.20/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 TPU v5p (per-chip, 95GB)?

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 TPU v5p (per-chip, 95GB) the right choice for OLMo 2 13B?

It works. Compare $4.75 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