Can I fine-tune LLMs with FrameTrain?

Yes. FrameTrain supports fine-tuning of Large Language Models (LLMs) via LoRA and QLoRA. You can import models like Llama, Mistral, or Phi directly from HuggingFace and adapt them to your own datasets.

Which GPUs are supported?

FrameTrain runs on all NVIDIA GPUs with CUDA support (from the GTX 10 series onward) as well as on Apple Silicon (M1/M2/M3/M4) via Metal Performance Shaders. Pure CPU-based training is also possible.

What's the difference compared to Google Colab or AWS SageMaker?

FrameTrain runs entirely locally on your machine. There are no ongoing cloud costs, your data is never uploaded, and you're not dependent on an internet connection. Ideal for sensitive or proprietary datasets.

Do I need programming knowledge?

No. FrameTrain is designed as a graphical application. You configure training parameters through an intuitive interface without writing any code. Advanced users can also use configuration files.

Does FrameTrain support LoRA training?

Yes, LoRA (Low-Rank Adaptation) is fully integrated. You can configure LoRA rank, alpha, and target modules directly in the app, which also enables fine-tuning on GPUs with limited VRAM.

Why does FrameTrain only cost €4.99/month?

FrameTrain is in Early Access. The price is intentionally low so that as many developers and researchers as possible can get access early. The price increases to €9.99 after the first 100 users – sign up now to lock in the lower price permanently.

Does FrameTrain run on Apple M1/M2/M3/M4?

Yes. FrameTrain natively supports Apple Silicon via Metal Performance Shaders (MPS). Models are accelerated on the M-chip's GPU without requiring CUDA.

Which model formats are supported?

FrameTrain supports PyTorch models (.pt, .bin), SafeTensors, and the GGUF format. Models can be imported directly from the HuggingFace Hub or loaded locally.

Can I use my own datasets?

Yes. You can load your own datasets in CSV, JSON, or JSONL format. FrameTrain also supports datasets directly from the HuggingFace Datasets Hub.

Does FrameTrain support mixed precision and gradient checkpointing?

Yes. FP16 and BF16 mixed-precision training are fully supported, as is gradient checkpointing for memory-efficient training on GPUs with limited VRAM.

Home GuidesGPU Guide

HardwareBuying Guide 2026

Best GPU for Local
ML Training & LLM Fine-Tuning

VRAM is the decisive factor. Here you'll find the right GPU for every budget and use case – from the affordable RTX 3060 to the Apple M4 Max.

Quick Summary: VRAM Rules of Thumb

6–8 GB

QLoRA on 7B models (just barely)

RTX 3070, RTX 4060

12–16 GB

LoRA on 7B, QLoRA on 13B

RTX 3060 12G, RTX 4070

24 GB+

LoRA on 13B, larger batches

RTX 3090, RTX 4090

RTX 4090

Best Choice

24 GB VRAMapprox. ~$1,950

Pros

Fastest consumer GPU
24 GB VRAM for 13B models
CUDA 8.9 – all features

Cons

✕Most expensive option
✕450W TDP – high power draw

Good for

13B models with QLoRA
7B models with full LoRA
Fast iteration cycles

RTX 4070 Ti / 4080

Recommended

12–16 GB VRAMapprox. ~$750–$1,200

Pros

Excellent performance/price ratio
Good for 7B models
Moderate power draw

Cons

✕13B models only with QLoRA
✕12 GB can get tight with large batches

Good for

7B models (LoRA & QLoRA)
Instruction tuning
NLP classification

RTX 3090 / 4070

Great Value

12–24 GB VRAMapprox. ~$650–$950

Pros

Excellent price-to-performance
RTX 3090: 24 GB VRAM
CUDA 8.6 – wide compatibility

Cons

✕RTX 3090: older architecture, fewer CUDA cores
✕RTX 4070: only 12 GB VRAM

Good for

Most 7B models without issues
LoRA & QLoRA fine-tuning

Apple M3 / M4 Pro & Max

Apple Alternative

18–128 GB Unified VRAMapprox. from ~$2,500 (Mac)

Pros

Very high unified memory bandwidth
M3/M4 Max: up to 128 GB
Efficient & quiet
Perfect for macOS users

Cons

✕More expensive than an equivalent NVIDIA setup
✕MPS slower than CUDA for some ops
✕No CUDA ecosystem

Good for

Local LLM inference
Fine-tuning via Metal (MPS)
Studio/creative workflows

RTX 3060 12GB

Entry Level

12 GB VRAMapprox. ~$270–$380

Pros

Cheapest entry point
12 GB VRAM for QLoRA
Broad CUDA support

Cons

✕Slower on larger models
✕No 7B LoRA without quantization

Good for

QLoRA on 7B models
Small classification models
Learning & experimenting

Your GPU Is Compatible – Get Started Now

FrameTrain supports all NVIDIA GPUs from the GTX 1060 (CUDA) and Apple Silicon M1–M4 (Metal). Starting at €4.99/month, cancel anytime.

Get Started

LoRA Guide

Best GPU for LocalML Training & LLM Fine-Tuning

Quick Summary: VRAM Rules of Thumb

RTX 4090

Pros

Cons

Good for

RTX 4070 Ti / 4080

Pros

Cons

Good for

RTX 3090 / 4070

Pros

Cons

Good for

Apple M3 / M4 Pro & Max

Pros

Cons

Good for

RTX 3060 12GB

Pros

Cons

Good for

Your GPU Is Compatible – Get Started Now

Best GPU for Local
ML Training & LLM Fine-Tuning