Can my PC run this model?
Whether a model runs on your machine comes down to two numbers you already have: how much memory the card holds, and how fast it can read it. One decides if the model loads, the other decides whether you will enjoy using it.
The short answer
Computed from this tool’s default settings — your hardware and the rest as most people start. Change them below for your own case.
RTX 4090 runs a 8B model at Q8_0, using about 10.9 GB of 22.1 GB usable and generating roughly 95 tokens per second.
Can My PC Run This Model?
A straight yes or no for any device in our catalogue, from a Raspberry Pi to an H100.
RTX 4090 runs a 8B model at Q8_0, using about 10.9 GB of 22.1 GB usable and generating roughly 95 tokens per second.
What each setting changes
Every input moves the result for a reason. This is what each one does and where to find the value for your own machine.
| Setting | Default | What it changes |
|---|---|---|
| Your hardware | RTX 4090 · 24 GB | The machine the model runs on. Usable memory decides what fits and memory bandwidth decides how fast it runs, so this moves every figure below. |
| Model size | 8B — Llama 3.1 8B class | Capacity follows total parameters, even for mixture-of-experts models. |
| Context length | 8192 tokens | The KV cache grows linearly with this. It is the biggest lever you have. |
Real answers across model size
The same calculation run at a range of settings, with everything else left at its default. These are computed by the tool itself, not written by hand.
| Model size | Verdict | Quantisation | Memory needed | Expected speed |
|---|---|---|---|---|
| 1B — Llama 3.2 1B class | Yes | Q8_0 | 2.55 GB | 759 tok/s |
| 7B — Mistral / Qwen 7B class | Yes | Q8_0 | 9.74 GB | 108 tok/s |
| 14B — Qwen 14B class | Yes | Q8_0 | 17.7 GB | 54.2 tok/s |
| 32B — Qwen 32B class | Yes | Q4_K_M | 21.8 GB | 44.8 tok/s |
| 120B — gpt-oss 120B class | No | 36.8 GB | MacBook Pro M4 Max |
RTX 4090 runs a 1B model at Q8_0, using about 2.55 GB of 22.1 GB usable and generating roughly 759 tokens per second.
How this is calculated
No lookup tables and no invented constants. Here is the arithmetic, so you can check it against your own numbers.
Memory for a model is three things added together: the weights, which are parameters × bits-per-weight ÷ 8; the KV cache, which grows linearly with context length; and about a gigabyte of runtime overhead for the CUDA context, activations and framework.
Capacity follows a model’s total parameter count even for mixture-of-experts designs, because the router may select any expert on the next token and all of them must stay resident. Only throughput follows the active count.
A model that does not fit entirely can still run by offloading layers to system RAM, but system memory is roughly an order of magnitude slower than VRAM, so throughput collapses. A smaller model that fits is almost always the better answer.
These are well-founded engineering estimates, not benchmark results. Your quantisation, runtime and context length all move the real number, and usable memory is an assumption rather than a specification. See the full methodology for every assumption behind these figures.
Who this is for
The three situations that bring people to this calculation.
New to local AI
Find out what your existing PC can already do, before buying anything.
Comparing options
See the largest model your machine handles at usable speed.
Troubleshooting
Understand why a model loaded but crawls.
How to use this calculator
Four steps, no account, nothing leaves your browser.
Set your hardware
Start at the top of the panel. Every figure recalculates as you change it — there is no submit button, because watching the number move is the point.
Adjust the rest to match your setup
2 further settings: model size, context length. Defaults are the common case, so change only what differs for you.
Read the headline, then the table
The large figure answers the question. The table underneath shows how the answer changes across nearby settings, which is usually where the decision actually gets made.
Check it against the method
The arithmetic is written out above. If a number looks wrong for your hardware, the assumptions are the first place to look — usable memory and quantisation are the two that vary most.
Can my PC run this model: common questions
The questions people ask about this, answered without hedging.
What are the minimum requirements to run an LLM locally?
An 8B model at Q4_K_M needs roughly 6 GB of memory, which puts it inside an 8GB card or a 16GB laptop running on CPU. Below that, 1B–3B models run on almost anything, including a Raspberry Pi, just slowly.
Do I need an NVIDIA GPU?
No. Apple Silicon runs local models well through Metal, and AMD cards work through ROCm and Vulkan. NVIDIA has the smoothest software story, but it is not a requirement.
How fast is fast enough?
About 15 tokens per second is a little faster than most people read, so anything at or above that feels immediate. Between 5 and 15 is usable, and below 5 is a batch job.
Can I use my GPU and CPU together?
Yes — llama.cpp splits layers between them. It works, but it is much slower than it sounds, because every token passes through every layer and the slow side sets the pace.
Related local AI tools
All 50 run on the same arithmetic, so answers across them agree.
More on will it fit
Context
Speed
Quantisation
Runtime & errors
Cost
Fine-tuning
Searches this page answers
Different ways of asking the same question, all resolved above.
Now find the models that fit
Sizing is only half the problem. Model Radar takes your hardware and shows which models actually run on it, ranked by what they are good at — the same arithmetic as this page, applied to every model worth running.