Tools + TOFU

How many GPUs do you need for AI inference?

A GPU capacity calculator estimates whether a model fits in memory and how many GPUs you need. Enter parameter count, quantization, context length and concurrent sessions; the tool adds weights, KV cache and runtime overhead to produce VRAM per replica, recommended GPU count and headroom, so you can plan capacity before buying or reserving hardware.

Key facts

Tool typeFree GPU and VRAM capacity estimator for inference
InputsParameter count, quantization, context length, concurrency, GPU memory
OutputsVRAM per replica, GPU count, utilization headroom
Memory componentsModel weights, KV cache, activations and runtime overhead
Quantization supportCommon bit widths from full precision to 4-bit
Companion toolsLocal model recommender and quantization calculator
Managed alternative30+ models hosted behind one API, no GPU operations
Product statusLive

TL;DR

  • Weights are only the first term: KV cache scales with context and concurrency.
  • Quantization cuts memory per weight but can change quality — validate on your evals.
  • Plan for peak concurrency, not average, and keep 15-20 percent headroom.
  • Multi-GPU adds communication overhead; not every model scales linearly.
  • Compare owning GPUs with the hosted API before committing to hardware.

How it works, step by step

  1. Identify the candidate model and its parameter count from the model card.
  2. Choose a quantization level and confirm it is supported by your inference runtime.
  3. Measure worst-case context length and expected concurrent sessions.
  4. Open the GPU capacity calculator and enter model, quantization, context and concurrency.
  5. Read VRAM per replica and the recommended GPU count, noting headroom.
  6. Adjust quantization or context and re-run to see how memory changes.
  7. Compare the cost of that hardware profile with the hosted API on the pricing page.
1Identify thecandidate model andits parameter count2Choose aquantization leveland confirm it is3Measure worst-casecontext length andexpected concurrent4Open the GPUcapacity calculatorand enter model,5Read VRAM perreplica and therecommended GPU6Adjust quantizationor context andre-run to see how

Try it yourself

Open the GPU capacity calculator →

The memory math that matters

Inference memory has three main parts. Model weights dominate at rest: a 70-billion-parameter model needs roughly 140 GB in 16-bit precision, halving at 8-bit and halving again near 4-bit. The KV cache grows with context length and the number of concurrent sequences, and it is often the term teams forget. Runtime overhead — activations, framework buffers and fragmentation — adds a further percentage. A capacity calculator combines these so the estimate reflects a serving replica, not a single idle model. Measure actual serving memory in a pilot before committing to a hardware bill of materials.

Quantization is a trade, not a free win

Lower precision buys memory and sometimes throughput, but it can shift output quality, especially on reasoning and structured-output tasks. The right process is mechanical: pick a quantization that fits your hardware budget, then run your own eval set at that precision and compare against the full-precision baseline. If accuracy holds, keep it; if it does not, move down a model size rather than up in bits. Remember that KV cache can often be quantized separately, which helps long-context workloads without touching weight precision.

Owning GPUs versus using an API

Hardware only makes sense at sustained, predictable utilization. Below that, the fixed cost of GPUs plus operations, monitoring, upgrades and idle time usually exceeds a managed API. Run both numbers: the calculator gives you the hardware profile and the pricing page shows flat monthly self-serve plans for 30+ hosted models. A common compromise is hybrid — keep sensitive or high-volume workloads on owned GPUs, and burst everything else to the API through the same OpenAI-compatible interface.

Honest comparison

Planning factorWhat to enterWhy it changes the answerCommon mistake
ParametersExact parameter count of the candidate modelWeights dominate base memoryUsing a similarly sized model's numbers
Quantization8-bit, 4-bit or full precisionHalves memory per step downAssuming no quality change
Context lengthWorst-case tokens per requestKV cache scales with contextSizing for average prompts
ConcurrencyPeak simultaneous sessionsEach sequence adds cachePlanning for average load
Runtime overheadFramework buffers and fragmentationAdds headroom above the mathPacking GPUs to 100 percent
GPU memoryMemory per card you plan to buy or rentDetermines GPUs per replicaIgnoring multi-GPU overhead

Frequently asked questions

How much VRAM does a model need?

It depends on parameter count, quantization, context length and concurrency. The calculator combines all four; weights alone are never the full answer for a serving system.

What is the KV cache?

The memory used to store attention keys and values for tokens in the context. It grows with context length and the number of concurrent sequences, and it is often the largest variable term.

Does 4-bit quantization always work?

No. It reduces memory but can change output quality. Validate on your own eval set at that precision before committing.

How much headroom should I leave?

Keep roughly 15-20 percent above the calculated requirement for fragmentation, bursts and framework overhead.

Can one GPU serve multiple models?

Sometimes, if their combined weights and caches fit with headroom. Otherwise route requests to separate replicas and load-balance.

Is multi-GPU always faster?

No. Communication overhead grows with GPU count, and small models may lose throughput. Measure per-replica latency before scaling out.

When is an API cheaper than GPUs?

At low or spiky utilization, almost always. Compare your hardware profile with hosted pricing before buying.

Does Plugsky host these models?

Yes. 30+ models are available behind one OpenAI-compatible API, so you can skip GPU operations entirely and see the live pricing page for plans.