Ask VRAMNeeded
Type your GPU (or VRAM), what you want to do and, if you like, a context length. Examples: "I have a 3060, what can I run for coding?", "24GB with 32k context", "Can a 4090 run gpt-oss-120b?".
Ask VRAMNeeded
Answers come from our own fit data in your browser (no AI model, nothing sent anywhere). Same rule as the calculator: weights + FP16 KV cache under 90% of memory.
How it works
- It reads your GPU or VRAM amount, a use case (coding, reasoning or general chat) and a context length (4K, 8K, 32K or 128K; 8K if you don't say).
- It checks every model in our list with the same rule as the calculator: real download sizes plus an FP16 KV cache, under 90% of memory.
- It runs entirely in your browser from a data file on this site. There's no AI model behind it, and your question is not sent anywhere.
Don't know your GPU? Scan it in the browser. Comparing costs? See the cheapest way to run each model.
Get fit alerts for your GPU
Join the fit-alerts waitlist
Alerts are not running yet. Join the waitlist and pick your card once; when alerts launch we will email you once to confirm, and only after you confirm will you get the first alerts. Unsubscribe any time.