How much VRAM do I need to fine-tune this model?
Fine-tuning used to mean a server. QLoRA changed that: by quantising the base model to 4-bit and training small adapters on top, the card that runs a model can usually train it.
The short answer
Computed from this tool’s default settings — model size and the rest as most people start. Change them below for your own case.
A 8B model QLoRA-fine-tunes in about 6.00 GB, which fits RTX 4090. 16-bit LoRA would need 22.0 GB and a full fine-tune 128.0 GB.
QLoRA VRAM Calculator
QLoRA, LoRA and full fine-tuning memory for any model size, against your card.
A 8B model QLoRA-fine-tunes in about 6.00 GB, which fits RTX 4090. 16-bit LoRA would need 22.0 GB and a full fine-tune 128.0 GB.
| Method | Memory | On your card |
|---|---|---|
| QLoRA (4-bit base) | 6.00 GB | Fits |
| LoRA (16-bit base) | 22.0 GB | Fits |
| Full fine-tune | 128.0 GB | Short by 105.9 GB |
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 |
|---|---|---|
| Model size | 8B | 9 options, from 1B to 405B. |
| 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. |
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 | Cheapest method that fits | Your memory | QLoRA needs | LoRA needs |
|---|---|---|---|---|
| 1B | QLoRA | 22.1 GB | 1.20 GB | 2.70 GB |
| 3B | QLoRA | 22.1 GB | 3.50 GB | 8.00 GB |
| 7B | QLoRA | 22.1 GB | 5.00 GB | 19.0 GB |
| 8B | QLoRA | 22.1 GB | 6.00 GB | 22.0 GB |
| 14B | QLoRA | 22.1 GB | 8.50 GB | 33.0 GB |
A 1B model QLoRA-fine-tunes in about 1.20 GB, which fits RTX 4090. 16-bit LoRA would need 2.70 GB and a full fine-tune 16.0 GB.
How this is calculated
No lookup tables and no invented constants. Here is the arithmetic, so you can check it against your own numbers.
QLoRA quantises the base model to 4-bit and trains small adapters on top, so the card that runs a model can usually train it too. LoRA keeps the base at 16-bit and trains the same adapters, costing roughly three to four times more memory. A full fine-tune updates every weight and carries gradients plus optimiser state — about 16 bytes per parameter.
The QLoRA and LoRA figures are Unsloth’s published requirements, interpolated between the sizes they list rather than extrapolated from a formula. They assume a short sequence length and a small batch; longer sequences and larger batches push the real number up.
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.
First fine-tune
Check your card can do it before setting anything up.
Choosing a base
Find the largest model you can actually train.
Renting
Size the instance you need rather than overpaying.
How to use this calculator
Four steps, no account, nothing leaves your browser.
Set model size
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
1 further setting: your hardware. 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.
How much VRAM do I need to fine-tune this model: common questions
The questions people ask about this, answered without hedging.
Can I fine-tune a 7B model on a 24GB card?
With QLoRA, comfortably — a 7B model needs about 5 GB. With 16-bit LoRA it needs roughly 19 GB, which fits but leaves little headroom for a longer sequence length. A full fine-tune of 7B needs around 112 GB and is not a single-card job.
How much VRAM do I need to fine-tune a 7B model?
About 5 GB with QLoRA, which fits comfortably on an 8GB card. Sixteen-bit LoRA needs roughly 19 GB, and a full fine-tune around 112 GB.
Can I fine-tune on a Mac?
Yes, through MLX, which handles LoRA on Apple Silicon natively. Unified memory means capacity is rarely the binding constraint; bandwidth is, so expect a run to take noticeably longer than on a discrete card of the same memory.
Does QLoRA hurt quality compared to LoRA?
Very little in practice. The base is quantised but frozen, and the adapters train in higher precision, so published comparisons find the gap small relative to the memory saved.
Do these numbers include the dataset?
No — they cover the model, gradients and optimiser state. Activations scale with batch size and sequence length, so leave 15–20% headroom on top.
Related local AI tools
All 50 run on the same arithmetic, so answers across them agree.
More on fine-tuning
Will it fit
Context
Speed
Quantisation
Runtime & errors
Cost
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.