Falcon 3 · TII
Fine-tuning Falcon 3 3B
What does it take to fine-tune Falcon 3 3B?
Falcon 3 3B needs 10 GB for half-precision LoRA training and 6 GB in four-bit, and 7 GB to serve. It supports supervised fine-tuning, preference tuning (dpo), under the TII Falcon LLM licence. The cheapest qualifying class is RTX 3080 at $0.09 per GPU-hour.
The second rung, and the smallest Falcon 3 that holds output format reliably across long generations. A base model rather than an instruction-tuned one, so expect to supply the conversational behaviour yourself in the training data.
What to know before choosing it
The smallest Falcon 3 that holds output format reliably across longer generations. Below this size the family, like most, starts dropping structure under pressure toward the end of a response.
Being a base model, expect the training data to carry the conversational behaviour as well as the task. That means more examples than an instruction-tuned model of the same size would need, and a cleaner result when you want a voice that is entirely yours.
Specification
| Parameters | 3B |
|---|---|
| Architecture | Dense |
| Size tier | Small |
| LoRA training memory | 10 GBHalf precision, frozen base |
| QLoRA training memory | 6 GBFour-bit base, higher-precision adapter |
| Serving memory | 7 GBHalf precision, before attention cache |
| Objectives | SFT, DPO |
| Licence | TII Falcon LLM |
| Repository | tiiuae/Falcon3-3B-Base |
What it costs to train
Every class with enough memory for four-bit training, cheapest first. Total cost is the rate multiplied by wall time, so the cheapest rate is not always the cheapest run — a faster class that finishes sooner frequently wins.
| GPU class | VRAM | Training / hr | Serving / hr | Fits |
|---|---|---|---|---|
| RTX 3080 | 12 GB | $0.09 | $0.11 | LoRA and QLoRA |
| RTX 4000 Ada | 20 GB | $0.09 | $0.11 | LoRA and QLoRA |
| L4 | 24 GB | $0.17 | $0.20 | LoRA and QLoRA |
| RTX 3090 | 24 GB | $0.22 | $0.26 | LoRA and QLoRA |
| RTX 4090 | 24 GB | $0.38 | $0.43 | LoRA and QLoRA |
| A40 | 48 GB | $0.42 | $0.48 | LoRA and QLoRA |
| RTX 6000 Ada | 48 GB | $0.61 | $0.71 | LoRA and QLoRA |
| A6000 | 48 GB | $0.65 | $0.75 | LoRA and QLoRA |
| L40S | 48 GB | $0.78 | $0.90 | LoRA and QLoRA |
| A100 40 GB | 40 GB | $1.17 | $1.35 | LoRA and QLoRA |
| A100 80 GB | 80 GB | $1.68 | $1.94 | LoRA and QLoRA |
| H100 80 GB | 80 GB | $2.59 | $2.99 | LoRA and QLoRA |
| H200 | 141 GB | $4.55 | $5.25 | LoRA and QLoRA |
Serving fits on RTX 3080 at $0.11 per hour — before the attention cache, which grows with context length and concurrency.
Good starting point for
- Building a specific conversational style from scratch
- Cost-sensitive generation
Preference tuning on this model
Preference tuning holds a frozen reference copy of the model alongside the one being trained, so budget roughly 20 GB rather than 10 GB. That is the thing that catches people out — supervised training on this model fits on a class that preference tuning will overflow.
Other sizes in this family
| Model | Params | QLoRA | Objectives |
|---|---|---|---|
| Falcon 3 1B | 1B | 3 GB | SFT, DPO |
| Falcon 3 7B | 7B | 12 GB | SFT, DPO |
| Falcon 3 10B | 10B | 16 GB | SFT, DPO |
Comparable sizes elsewhere
Frequently asked questions
How much VRAM does it take to fine-tune Falcon 3 3B?
10 GB for half-precision LoRA and 6 GB for four-bit QLoRA. Serving needs 7 GB. Preference tuning roughly doubles the training figure, because a frozen reference model is held alongside the one being trained.
What is the cheapest way to fine-tune Falcon 3 3B?
Four-bit QLoRA on RTX 3080 at $0.09 per GPU-hour is the cheapest class that meets the 6 GB threshold. Note that quantised training is slower per step, so a faster class sometimes costs less over the whole run.
Can I download the weights after fine-tuning Falcon 3 3B?
Yes. Every finished run exposes its trained weights for download, and publishing to a model hub is a single call with a generated model card recording the base model and version the adapter applies to.
What licence does Falcon 3 3B carry?
TII Falcon LLM. The licence follows the fine-tune — a derivative inherits the base model’s terms, and those terms pass to anyone you give the model to.
Last verified 6 August 2026. Memory thresholds are the platform's own admission limits.
Fine-tune Falcon 3 3B
Upload a dataset, forecast the run, and see the cost before any compute is leased.