Cerebras44 GB VRAM20,000 GB/s

Phi-4 14B on Cerebras CS-3 (WSE-3): Does It Fit, How Fast, What It Costs

Phi-4 14B at FP16 needs 28.0 GB for its weights and 38.7 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
38.7 GB needed / 44 GB available
Decode speed
464.3
tok/s per stream
Aggregate
5571
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 GB38.7 GB464.35571399.2 ms—✓
2× Cerebras88 GB38.7 GB789.318943214.8 ms—✓
4× Cerebras176 GB38.7 GB157975771122.4 ms—✓
8× Cerebras352 GB38.7 GB315730308676.2 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?
4K23687.2 MB✓ fits
8K111.34 GB✓ fits
33K25.37 GB✓ fits
66K110.74 GB✓ fits
131K021.47 GB✗ does not fit
Other GPUs that run Phi-4 14B
Same workload, same quantization — ranked by value (tokens per dollar-hour).
GPUGPUsAggregate tok/sCost / 1M tok
RTX 5090 32GB2×1697$0.1637
AMD MI300X 192GB1×1476$0.4704
A100 40GB SXM41×431.8$0.5790
AMD MI325X 288GB1×1671$0.5817
A100 80GB SXM41×557.1$0.5983
Intel Gaudi 3 128GB1×919.3$0.6043
Frequently asked questions

Can Phi-4 14B run on Cerebras CS-3 (WSE-3)?

Yes. Phi-4 14B at FP16 needs 38.7 GB including the KV cache, and Cerebras CS-3 (WSE-3) has 44 GB — so it fits on a single GPU.

How fast is Phi-4 14B on Cerebras CS-3 (WSE-3)?

464.3 tokens/sec per stream and 5571 tok/s aggregate with continuous batching at 8 concurrent users. Time to first token is 399.2 ms, and inter-token latency is 2.2 ms.

What does Phi-4 14B 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 Phi-4 14B need on Cerebras CS-3 (WSE-3)?

28.0 GB for the weights at FP16, plus 10.7 GB of KV cache for 8 concurrent requests at 8K context — 38.7 GB total. Every extra concurrent user at long context costs more, so leave headroom.

Is Cerebras CS-3 (WSE-3) the right choice for Phi-4 14B?

It works. Compare — per 1M tokens against the alternatives below — the cheapest option that fits the same workload is 4× RTX 3080 10GB at $0.0772, and the fastest is 1× Cerebras CS-3 (WSE-3) at 5571 tok/s.

Adjust the workload yourself

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

Open in calculator