The UK AI Toolkit 2026

Honest reviews, free tier status, and UK pricing

What you actually pay — in GBP

Subscription vs API — which is right for you?

What to use for your specific job

33 terms explained in plain English

The basics

Artificial Intelligence (AI)

Computer systems that can perform tasks that normally require human intelligence — understanding language, generating text, recognising images, and making decisions.

Large Language Model (LLM)

An AI trained on vast amounts of text that can understand and generate human-like writing. GPT-4, Claude, and Gemini are all LLMs.

Generative AI

AI that creates new content — text, images, audio, or video — rather than just classifying or analysing existing content.

Chatbot

An AI you can have a conversation with. ChatGPT, Claude, and Gemini are examples. Not all chatbots are AI — older ones run on fixed scripts.

AI Agent

An AI system that can take actions — browsing the web, running code, sending emails — rather than just generating text responses.

Multimodal

An AI that can handle multiple types of input: text, images, audio, and sometimes video. GPT-4o and Claude Sonnet are multimodal models.

Foundation Model

A large AI model trained on broad data, designed to be adapted for many tasks. GPT-4 and Claude are foundation models.

Output

The text, image, or other content an AI produces in response to your input. The quality of output depends heavily on the quality of your prompt.

Iteration

The process of refining AI outputs through multiple attempts. Most effective AI use involves asking, reviewing, and asking again until you get what you need.

Open Source

AI models whose code and weights are publicly available. Examples: Llama, Mistral. Can be run locally or on cheaper cloud infrastructure.

Technical terms

Token

The smallest unit of text an AI processes. Roughly, 1 token ≈ 0.75 words in English. API pricing is charged per 1,000 or 1,000,000 tokens.

Context Window

The maximum amount of text an AI can see at once — your input plus its output combined. Larger context windows let you work with longer documents.

Transformer

The neural network architecture that powers most modern AI. Introduced by Google in 2017, it underlies ChatGPT, Claude, and Gemini.

Diffusion Model

The type of AI used to generate images (Midjourney, DALL-E, Stable Diffusion). It works by learning to remove noise from randomised images step by step.

Embedding

A way of converting text into numbers that capture meaning. Used to power semantic search and AI comparison tools.

Fine-tuning

Training a pre-built AI model further on your own data to specialise it — for example, training it on your company's writing style or product knowledge.

RAG (Retrieval-Augmented Generation)

A technique where an AI looks up information from a specific database before responding. Reduces hallucinations and keeps information current.

Inference

The process of an AI generating a response. When you use an AI API, you are paying for inference compute — the cost of running the model.

API (Application Programming Interface)

A way for software to connect to an AI service programmatically. Using the API is typically much cheaper than a subscription but requires technical setup.

Temperature

A setting that controls how predictable or creative an AI's responses are. Low (0) = consistent and reliable. High (1+) = more varied and creative.

Prompting & usage

Prompt

The instruction or question you give to an AI. The quality of your prompt directly determines the quality of the response.

Prompt Engineering

The practice of writing prompts that reliably produce useful AI outputs. Good prompts are specific, provide context, and state the desired format.

System Prompt

A hidden instruction given to an AI at the start of a session, defining its role, tone, or constraints. Often used in business applications.

Zero-Shot

Asking an AI to complete a task without giving examples. Works for straightforward tasks; use few-shot when format precision matters.

Few-Shot

Giving an AI a few examples of what you want before asking it to do the task. Usually produces better results for specific formats or styles.

Hallucination

When an AI generates convincingly wrong information — presenting false facts confidently. Always verify AI output when accuracy matters.

Pricing & performance

Input Tokens

The tokens in the text you send to an AI. Input is usually priced cheaper than output. A 300-word prompt is approximately 400 input tokens.

Output Tokens

The tokens in the text an AI sends back. Output is priced higher than input on most APIs because it requires more compute to generate.

Benchmark

A standardised test used to compare AI model performance — for example, MMLU (general knowledge) or HumanEval (coding). Useful for comparing models technically.

Legal & compliance (UK)

GDPR

The UK and EU's data protection regulation. When using AI tools with customer data, ensure personal data is processed lawfully. Check where each tool stores your data.

ICO

The Information Commissioner's Office — the UK's data protection regulator. Businesses collecting personal data must register with the ICO (£40/year for small businesses).

AI Safety

The field focused on ensuring AI systems behave as intended and do not cause unintended harm. The UK's AI Safety Institute leads national policy in this area.

Data Residency

The physical location where your data is stored. For UK GDPR compliance, confirm whether an AI tool stores data in the UK or EU.