Runyard is a free hardware-aware AI model browser. You enter your CPU, GPU, and VRAM and it instantly shows every local LLM that will run on your machine, ranked by speed and quality.

How much VRAM do I need to run local LLMs?

8GB of VRAM runs 7B models like Llama 3.1 8B and Mistral 7B at Q4 quantization. 16GB unlocks 13B models. 24GB lets you run Mixtral 8x7B and Llama 3 70B at lower quantization.

What is the best local LLM for my GPU?

Use Runyard at www.runyard.dev — enter your GPU and VRAM and the Model Radar will rank every compatible LLM for your exact hardware, showing estimated tokens per second for each model.

Can I run Llama 3 locally?

Yes. Llama 3.1 8B at Q4 runs on any 8GB VRAM GPU. Llama 3.1 70B needs around 40GB VRAM at Q4, or an Apple Silicon Mac with 64GB+ unified memory.

All tools

Cost · free, no sign-up

Is RAG or fine-tuning cheaper for my use case?

Fine-tuning is paid once; retrieval is paid on every request. That difference in shape, not the headline prices, is what decides which is cheaper for you.

Open the calculator Find models for your hardware All 50 tools

5 inputs4 questions answeredUpdated for 2026 hardware

Answer first

The short answer

Computed from this tool’s default settings — requests per day and the rest as most people start. Change them below for your own case.

Fine-tune pays back after0.6 days

At 1,000 requests a day, the retrieved context costs more than the fine-tune within 0.6 days. Fine-tuning is the cheaper shape here — but only if the knowledge is stable enough not to need redoing.

The calculator

RAG vs Fine-Tune Cost

A one-off training cost against the per-request cost of retrieval.

Fine-tune pays back after0.6 days

At 1,000 requests a day, the retrieved context costs more than the fine-tune within 0.6 days. Fine-tuning is the cheaper shape here — but only if the knowledge is stable enough not to need redoing.

Fine-tune once$5.52

RAG per request$0.00900

RAG per month$273.60

Break-even613 requests

Inputs

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
Requests per day	`1000`	Your own figure, starting from 1,000. Change it to match what you actually run.
Retrieved context per request	`3000 tokens`	RAG pays this on every single request. Fine-tuning pays once.
Input price	`3 per 1M`	Your own figure in per 1M, starting from 3. Change it to match what you actually run.
Fine-tune GPU hours	`8 hours`	Your own figure in hours, starting from 8. Change it to match what you actually run.
GPU hourly rate	`0.69 per hour`	Your own figure in per hour, starting from 0.69. Change it to match what you actually run.

Method

How this is calculated

No lookup tables and no invented constants. Here is the arithmetic, so you can check it against your own numbers.

The two cost shapes are different in kind: fine-tuning is a fixed cost paid once, RAG is a recurring cost paid on every request as extra input tokens. So the answer is a break-even in requests, not a verdict.

Cost is not the only axis, and often not the deciding one. RAG updates the moment your documents change and can cite them; a fine-tune bakes knowledge in and has to be redone. Size the money here, then decide on freshness.

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.

Use cases

Who this is for

The three situations that bring people to this calculation.

Use case 01

Choosing an approach

Find the request volume where fine-tuning wins.

Use case 02

High traffic

See what retrieved context costs at scale.

Use case 03

Planning

Weigh a one-off training cost against a recurring one.

Walkthrough

How to use this calculator

Four steps, no account, nothing leaves your browser.

Set requests per day
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
4 further settings: retrieved context per request, input price, fine-tune gpu hours, gpu hourly rate. 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.

Questions

Is RAG or fine-tuning cheaper for my use case: common questions

The questions people ask about this, answered without hedging.

Is RAG or fine-tuning cheaper?

At low volume, RAG — a fine-tune has to be paid for before it saves anything. At high volume the per-request context cost overtakes it, sometimes within days.

Which should I choose if cost is close?

RAG, in most cases. It updates the moment your documents change, it can cite sources, and it does not need redoing when the base model is replaced.

When is fine-tuning clearly right?

For behaviour rather than knowledge — tone, format, following a house style, or a task the base model does badly. Facts belong in retrieval; behaviour belongs in weights.

Can I use both?

Frequently the best answer: fine-tune for the format and behaviour you want, retrieve for the facts. They solve different problems.

The rest of the set

All 50 run on the same arithmetic, so answers across them agree.

Will it fit

Context

Speed

Memory

Quantisation

Runtime & errors

Fine-tuning

Hardware

Coverage

Searches this page answers

Different ways of asking the same question, all resolved above.

rag vs fine tuning costrag or fine tunewhen to fine tune vs ragfine tuning cost calculatorrag cost per requestretrieval vs training costrag vs fine tuning comparisonis rag or fine-tuning cheaper for my use caserag vs fine-tune costrag vs fine-tune cost onlinefree rag vs fine-tune cost

Next step

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.

Open Model Radar Browse all 50 tools