LLM VRAM calculatorCalculators

LLM VRAM calculator. Will it fit on your GPU?

Pick a model or type its size, the weight format (FP16, INT8, Q4 and more), context length and batch. See the weights, the KV cache, an overhead share you set and which common GPUs hold the total. An estimate from a stated formula. Nothing leaves your browser.

Cost to you

Free

Runs in your browser. No signup, no API call, no limit per hour.

browser
runs in
none
API calls
none
account

Use it now

Use the LLM VRAM calculator here

It runs in your browser. Nothing you type is sent anywhere, there is no signup and no limit.

Loading the tool.

Run it from your agent

Copy one prompt, give it to your agent.

Fill in the placeholder, paste it, and the agent runs the same job with current data on your looot account and shows the cost.

  • You give it: parameters, weight format, context length, batch size and the model's layer shape.
  • Billed per call, nothing while your agent is idle.
  • The agent shows the price before it calls.
Prompt: LLM VRAM calculator
Use looot to check if I can run [model] on my [GPU] at [context] tokens. 1. Set up https://looot.ai/skill.md. 2. Read the model's parameter count, layers, KV heads and head size from its config. 3. Work out weights plus KV cache for FP16, INT8 and Q4. 4. Say which format fits, and compare with the price of running the same model through a chat completion endpoint in the catalog.

Give this to your agent
Setup command
set up looot - https://looot.ai/skill.md

What comes back

What your agent gets from vram calculator

Field names as the providers return them. The exact set varies by endpoint, and your agent reads it from the catalog.

  • total GPU memory in GB
  • weights in GB
  • KV cache in GB
  • which GPUs fit and how many cards

Use it for

What people do with vram calculator

  • Buy the right GPU

    See how much memory a model needs before you rent or buy a card.

  • Pick a quantization

    Compare FP16, INT8 and Q4 for the same model and context.

  • Run models locally

    Check if an open model such as a Gemma 4 fine-tune fits your machine.

  • Size long context

    See how the KV cache grows as you raise the context length or the batch.

Set up once

Three steps. Your agent does the rest.

Keep going

More free tools

FAQ

Questions about vram calculator

Weights take parameters times bits per weight divided by 8. The KV cache takes 2 times layers times KV heads times head size times context times batch times bytes per value. Add an overhead share for the runtime. The calculator shows each part.

Pick Qwen2.5 7B Instruct and switch the weight format. The weights alone are about 15 GB at FP16 and about 4.3 GB at Q4_K, before the KV cache and overhead.

The presets use the parameter count and the config.json of each model on Hugging Face, read 2026-10-14. Choose Custom to type any model's numbers.

The calculator counts every layer at full context. Models with sliding window layers, such as the Gemma family, keep a shorter cache in most layers, so they use less.

Humanizer by jialinyyzz is a Gemma 4 12B fine-tune with about 11.9 billion parameters by its GGUF header. Choose Custom, type 11.9 and a weight format to see the weights. Read its layer numbers from the model card for the cache.

It depends on the runtime, so the field is yours to set. Start at the default and check against the memory your runtime reports.

Yes. It runs in your browser, needs no account and has no limit.
For agents
tool
LLM VRAM calculator
category
Calculators
input
Parameters, weight format, context length, batch size and the model's layer shape
output
total GPU memory in GB, weights in GB, KV cache in GB, which GPUs fit and how many cards
price
per call, read live from the catalog on the web page
web tool
runs in the visitor's browser, no API call, free and unlimited
schedule
on demand
setup
https://looot.ai/skill.md
llms.txt
https://looot.ai/llms.txt
updated
2026-10-14
markdown /free-tools/vram-calculator.md

Run vram calculator from your own agent.

Start your workspace, top up, and paste the prompt into the agent you already use. It runs the job with live data and shows the cost first.

Start your workspace See all free tools