👋Welcome

New to AI models?

There are hundreds of AI models out there — each built by different companies, with different strengths, prices, and use cases. This page helps you understand what they are and which ones to consider.

What is an AI language model?

An AI language model is a system trained on vast amounts of text that can understand and generate human language. You describe what you want in plain English (or any language), and the model responds — writing, explaining, coding, summarising, or reasoning through a problem.

Unlike search engines that retrieve existing content, AI models generate new responses tailored to your exact question. Think of them as a highly knowledgeable assistant that never tires, works 24/7, and can handle almost any written task you throw at it.

The key difference between models is capability, speed, and cost. A model like GPT-5.4 or Claude Opus 4.7 is extremely capable but costs more per query. Models like Gemini 2.5 Flash or GPT-4.1 Nano are fast, cheap, and perfect for high-volume tasks.

What do you want to do?

Pick a use case to see which models are recommended.

✍️

Writing & content

Drafting emails, blog posts, marketing copy, or creative writing.

💻

Coding & development

Writing code, debugging, code review, and architecture decisions.

🔬

Research & analysis

Summarising papers, analysing data, deep reasoning tasks.

🗣️

Conversations & chat

General questions, explanations, language learning, brainstorming.

🖼️

Image understanding

Describing images, reading charts, visual Q&A.

🤖

Building AI apps

Creating your own AI-powered product, API integrations, agents.

The major providers

These are the companies building the most widely used models. Follow them to get notified when they release something new.

Key concepts explained

Context window
How much text the model can "see" at once. A 200K context window means you can feed in roughly 150,000 words — about two full novels — in a single conversation.
Input / Output price
API costs are measured per million tokens (≈750,000 words). Input tokens are what you send; output tokens are what the model writes back. Output is usually 3–5× more expensive.
Benchmark
A standardised test used to compare models — like MMLU for general knowledge or SWE-bench for coding. Higher is better, but real-world performance varies by task.
Open-weights
Models like Llama 4 or Mistral where the actual model weights are freely downloadable. You can run them yourself, fine-tune them, or use them via a hosting provider.
Tool calling
The ability for a model to call external functions or APIs — for example, searching the web, querying a database, or running code.
Reasoning model
Models like o3 or Claude 3.7 Sonnet that "think out loud" internally before answering. Slower and more expensive, but much better at complex math and logic.

Ready to explore?

Create a free account to follow providers, get notified about new releases, and compare models side-by-side.