Top news
Investing in the public benefit OfficialAnthropic NewsOct 18Popular Musk Pledges SpaceX Shares to ‘Every Kid in America’ WebBloomberg TechnologyOct 19Popular Websites are closing their doors to AI agents WebThe Verge AIOct 17 This is the new ChatGPT Voice, powered by GPT Live OfficialOpenAI on YouTubeOct 18 How AI giants will reshape global philanthropy WebAxios AIOct 19 Gemini 4 Argon Really Surprised Me WebTwo Minute PapersOct 19 Google Gemini 4 agents just beat Minecraft with no human intervention WebEngadget AIOct 19 Who could it be? 🧐 OfficialGoogle on YouTubeOct 18 Anthropic gives outside AI watchdog “employee-like” access WebArs Technica AIOct 18 What are self-replicating ‘AI worms’, and why is Anthropic worried about them? WebThe Guardian AIOct 18
GlossaryTrainingIntermediate

Fine tuning

Also written: fine-tune, finetuning, SFT, Supervised Fine Tuning, Fine-tune model behavior, Malicious fine-tuning, MFT, Fine-tuned models

Adapting a general AI model to perform better on specific tasks or industries by training it on relevant data.

In plain terms

Fine tuning means taking a large, general-purpose AI model and training it a bit more on a specific type of data. Think of it like a chef who knows how to cook many things but then practices making only Italian food to become an expert Italian chef.

This extra training helps the AI model become much better at a particular job, like answering questions about a specific company's products or writing in a certain style. For example, a company might fine tune a model with its customer service logs to help answer customer questions more accurately.

The expert version

Fine tuning involves further training a pre-trained machine learning model, typically a large language model (LLM), on a smaller, task-specific dataset. This process adjusts the model's existing parameters to improve its performance on a particular domain or task, without retraining from scratch. It's a way to specialize a general model, making it more efficient and accurate for niche applications.

Tradeoffs include the cost and time for additional training, the risk of 'catastrophic forgetting' where the model loses some general capabilities, and the need for high-quality, relevant data. The effectiveness depends heavily on the quality and quantity of the fine-tuning dataset and the similarity of the target task to the original training data.

Why it matters now

Fine-tuning is crucial for adapting large, general-purpose AI models to specialized tasks and industries. Recent news shows companies like Anthropic and Meta are enabling developers to customize models for specific use cases, from telecommunications to image generation. This trend highlights the growing need for tailored AI solutions beyond broad capabilities.

Where it shows up 37 items, newest first
How each company uses it
Anthropic Claude
Anthropic is fine-tuning its Claude models to improve performance on specific industry use cases, such as in the telecommunications sector through partnerships.
Google Gemini
Google's Gemma models, with their widespread downloads, are likely being fine-tuned by developers to create specialized applications.
Meta Meta AI
Meta AI provides resources like the Llama Cookbook to help developers fine-tune their models, including for image-related tasks, expanding their capabilities.
Related terms

Large Language ModelsPre-trainingTransfer LearningPrompt EngineeringModel AdaptationDomain Specific Models

Often confused with
Pre-training
Pre-training involves training a model from scratch on a massive d
Prompt Engineering
Prompt engineering focuses on crafting effective input promp
Reinforcement Learning from Human Feedback (RLHF)
RLHF is a training technique
Origin

The concept of fine tuning in machine learning dates back to the late 1990s and early 2000s, with early work in areas like neural network adaptation. It became particularly prominent with the rise of deep learning and large pre-trained models. Source ↗

Try this about 5 minutes

You can experiment with fine tuning by using platforms that offer simplified interfaces for adapting models. For instance, some services allow you to upload a small dataset of text examples and generate a custom chatbot that responds in a specific style or answers questions about a particular topic.

Try it: Playbook entries that use thisPlaybook →

The Playbook: reusable how-tos, prompts, configs and tips pulled out of the news.

Questions people ask

How is fine tuning different from training a model from scratch?
Fine tuning starts with a model that has already been trained on a vast amount of data. It then continues training on a smaller, specific dataset to specialize the model, which is much faster and cheaper than training from scratch.
What kind of data do I need for fine tuning?
You need a dataset that is relevant to the specific task or domain you want the AI model to excel at. This could be question-answer pairs, text in a particular style, or examples of desired outputs.
Can fine tuning make an AI model biased?
Yes, if the fine tuning dataset contains biases, the model can learn and amplify them. It's crucial to use diverse and representative data during the fine tuning process.

First seen on Subvolts May 28, 2020. Updated Oct 12, 2026. Written by an AI model from the items above and reviewed against them weekly; report an error.