Google95 GB VRAM819 GB/s$4.2/hr

Qwen 3 30B-A3B (MoE) on TPU v5p (per-chip, 95GB): Does It Fit, How Fast, What It Costs

Qwen 3 30B-A3B (MoE) at FP8 (E4M3) needs 30.0 GB for its weights and 36.4 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
36.4 GB needed / 95 GB available
Decode speed
266.2
tok/s per stream
Aggregate
3194
tok/s @ 8 concurrent
Cost / 1M tokens
$0.3653
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 GB36.4 GB266.23194247.9 ms$0.3653✓
2× TPU190 GB36.4 GB452.510860139.3 ms$0.2149✓
4× TPU380 GB36.4 GB905.04344084.6 ms$0.1074✓
8× TPU760 GB36.4 GB181017375957.3 ms$0.0537✓
Concurrency vs context length
On 1× TPU v5p (per-chip, 95GB) at FP8 (E4M3). KV cache scales linearly with context, so this is the number of users your GPUs can actually serve.
ContextConcurrent usersKV per requestFits?
4K161412.3 MB✓ fits
8K80824.6 MB✓ fits
33K203.22 GB✓ fits
66K106.44 GB✓ fits
131K512.88 GB✓ fits
Other GPUs that run Qwen 3 30B-A3B (MoE)
Same workload, same quantization — ranked by value (tokens per dollar-hour).
GPUGPUsAggregate tok/sCost / 1M tok
RTX 5090 32GB2×23762$0.0117
AMD MI300X 192GB1×20670$0.0336
A100 40GB SXM41×6045$0.0414
AMD MI325X 288GB1×23400$0.0415
A100 80GB SXM41×7800$0.0427
Intel Gaudi 3 128GB1×12870$0.0432
Frequently asked questions

Can Qwen 3 30B-A3B (MoE) run on TPU v5p (per-chip, 95GB)?

Yes. Qwen 3 30B-A3B (MoE) at FP8 (E4M3) needs 36.4 GB including the KV cache, and TPU v5p (per-chip, 95GB) has 95 GB — so it fits on a single GPU.

How fast is Qwen 3 30B-A3B (MoE) on TPU v5p (per-chip, 95GB)?

266.2 tokens/sec per stream and 3194 tok/s aggregate with continuous batching at 8 concurrent users. Time to first token is 247.9 ms, and inter-token latency is 3.8 ms.

What does Qwen 3 30B-A3B (MoE) on TPU v5p (per-chip, 95GB) cost per million tokens?

$0.3653 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 Qwen 3 30B-A3B (MoE) need on TPU v5p (per-chip, 95GB)?

30.0 GB for the weights at FP8 (E4M3), plus 6.4 GB of KV cache for 8 concurrent requests at 8K context — 36.4 GB total. Every extra concurrent user at long context costs more, so leave headroom.

Is TPU v5p (per-chip, 95GB) the right choice for Qwen 3 30B-A3B (MoE)?

It works. Compare $0.3653 per 1M tokens against the alternatives below — the cheapest option that fits the same workload is 4× RTX 3080 10GB at $0.005513, and the fastest is 1× Cerebras CS-3 (WSE-3) at 78000 tok/s.

Adjust the workload yourself

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

Open in calculator