TuneSalon AI
Resources

Frequently Asked Questions

Common questions about TuneSalon AI, fine-tuning, and how to get started.

Fine-tuning is teaching an existing AI model new knowledge or behaviour using your own data. Instead of building a model from scratch, you start with a powerful base model and customise it to think, write, or respond the way you want. Read more in our full guide: What is Fine-tuning?

Anywhere you need AI to go beyond generic responses. Customer support that knows your products, writing that matches your voice, legal analysis in your firm's style, translations with your preferred terminology. Read more: Where Fine-tuning Matters

Start by preparing your training data. You can bring your own dataset or use our Dataset Generator to create one from your documents. Then pick a base model from our curated list, hit train, and wait. TuneSalon handles all the technical setup. When training is done, you get free chat time to start using your model immediately. For a full walkthrough, see our Tutorial.

Every new account receives 520 welcome credits, no payment required. That's enough for a full first experience: generate around 100 training examples with the Dataset Generator, train a 7B parameter model, and test it in chat with 20 minutes of free conversation included. After that, credit packs start from $5.

You have three options. Chat with it on TuneSalon — your training includes free chat time, so you can start using your model right away. Use it locally through the desktop app for complete privacy. Or export it as a GGUF file, which is a standalone format you can run with any compatible tool.

On the website, your data passes through our cloud servers for training and chat. If privacy is a priority, the TuneSalon desktop app lets you train and chat entirely on your own hardware. Nothing leaves your computer. Same interface, full offline capability.

TuneSalon offers a curated list of permissively licensed models: Qwen3-4B, Mistral-7B, Qwen3-8B, Ministral-8B, Qwen3-14B, Mistral-Small (24B), and with an H200 GPU tier, Qwen3-32B and Qwen3.5-35B. All are Apache 2.0 licensed models you can use freely, including commercially.

Yes. You can upgrade, downgrade, or cancel your subscription at any time. If you cancel or downgrade, your current tier stays active until the end of your billing period. Upgrades take effect immediately.

TuneSalon started from a simple belief: fine-tuning should be accessible to everyone, not just developers. That vision hasn't changed, and there's a lot more to come. Support for more model architectures, mixture-of-experts fine-tuning, agentic behaviour training, and whatever new techniques emerge in this fast-moving field. The goal is to keep growing with the AI landscape and make sure the barrier stays low, no matter how advanced the technology gets.