Microsoft · 14.7B parameters · MIT
What GPU do you need to run Phi-4 14B?Memory, cards and speed.
Phi-4 14B needs about 10 GB of GPU memory (VRAM) at its smallest (4-bit), so a 16 GB card such as an RTX 4080 runs it. The cheapest card on our list that fits, an RTX 4060 Ti 16GB, costs about $0.14 an hour. At its largest (16-bit) it needs about 31 GB.
GPU memory
Phi-4 14B, by version.
Includes about 1.5 GB for the program and a short conversation (about 4,000 tokens). Long conversations need more. Speeds are ceilings: real speeds are usually half to two-thirds of them. How we work it out
| Version | Memory needed | Card size | Cards that fit | From | Top speed |
|---|---|---|---|---|---|
| 4-bit | about 10 GB | 16–20 GB | RTX 4090, RTX 5090, RTX 3090, RTX 4080, RTX 3090 Ti, RTX 5080, RTX 5070 Ti, RTX 4070 Ti SUPER, RTX 5060 Ti 16GB, RTX 4060 Ti 16GB | $0.14/hr on an RTX 4060 Ti 16GB | about 203 tokens/s on an RTX 5090 |
| 8-bit | about 17 GB | 24 GB | RTX 4090, RTX 5090, RTX 3090, RTX 3090 Ti | $0.155/hr on an RTX 3090 | about 115 tokens/s on an RTX 5090 |
| 16-bit | about 31 GB | 32 GB and up | RTX 5090 | $0.524/hr on an RTX 5090 | about 61 tokens/s on an RTX 5090 |
Strong at reasoning and maths for its size. MIT license.
Run it
Start Phi-4 14B in minutes.
On KernelMachine, Rent a Chat model with it: it loads into the chat app by itself once the rental is ready, straight from Ollama to the machine, and is in the app's model menu. On any machine with Ollama, one command downloads and runs it. Ollama on a rented GPU
$ ollama run phi4:14b
# Model files: huggingface.co/microsoft/phi-4
Common questions
Questions, answered plainly.
How much VRAM does Phi-4 14B need?
4-bit: about 10 GB; 8-bit: about 17 GB; 16-bit: about 31 GB. That includes about 1.5 GB for the program and a short conversation; longer conversations need more.
Can an RTX 4090 run Phi-4 14B?
Yes. Phi-4 14B (8-bit) needs about 17 GB, which fits in the RTX 4090's 24 GB.
How fast does Phi-4 14B run?
At 4-bit, the ceiling is about 203 tokens/s on an RTX 5090: each word means reading the whole model from memory once. Real speeds are usually half to two-thirds of that.
Can I run Phi-4 14B on KernelMachine?
Yes. Rent a Chat model with Phi-4 14B, on this page, starts the chat app on the cheapest card with 16 GB or more, and phi4:14b loads into it by itself once it's ready. Already renting a Chat model? Pick Phi-4 14B on the rental's page, or download phi4:14b from the chat app's model menu.
Other models
Compare with other chat models.
Preparing an invite-only beta
Get an email when access opens.
No launch date is set. We'll email you once when KernelMachine opens. Nothing else.