Models · Learn
Models, explained
The leaderboard tells you the numbers. This tells you what they mean. Three short guides to what is different between the models, which one to use for your job, and how to run several of them together without the headache.
Updated
1,342 models in the live catalogue
What is different between the models
Claude, GPT, Gemini, Grok, and the open-weight models (DeepSeek, Qwen, Kimi, GLM, Llama, Mistral). Not specs, temperament: what each family is genuinely good at and when to reach for it.
6 min read Read DecideWhich model should you use?
A task-by-task router. Coding, autonomous agents, high-stakes reasoning, high-volume work, long documents. One question decides most of it, with live picks for every job.
5 min read Read Run itHow to run many models, not one
The best model changes every week. The operators who win route each step to the right one instead of betting on a single model. How routing works, and what it takes to run in production.
6 min read ReadReading, backed by real numbers
Every guide points back to the live catalogue. Pricing and benchmarks refresh from OpenRouter and Artificial Analysis, so the picks stay current as the frontier moves.
Live ranking board
Sort every model by intelligence, coding, agentic, speed, price, and context.
Leaderboard, explained
What the OpenRouter usage rank measures, and how it differs from a quality score.
Category boards
The current best models by job: coding, agentic, reasoning, and more.
Compare A vs B
Two models side by side on price, context, modalities, and licence.
The value frontier
The models that give the most capability for the least money.
Model pulse
What changed this week: new models, price moves, context expansions.
Common questions
What is the difference between AI models?
The frontier models are close on most single tasks, so the real differences are temperament and shape: how closely a model follows instructions, how it writes, how much it can read at once, how it behaves in an agent loop, and what it costs to run. Open-weight models add the option to run the model yourself at lower cost. The model families guide breaks down each one.
How do I know which AI model to use?
Start with the job, not the model. Name the task, weigh the cost of being wrong, then pick the cheapest model that clears the quality bar. The how-to-choose guide walks through this task by task with live picks for coding, agents, reasoning, high-volume work, and long documents.
Should a business use one model or several?
Several, for almost any real workflow. Different steps have different needs, so matching each step to the right model is cheaper and better than forcing one model to do everything. That practice is called model routing, and it is how the intelligence layer is actually run in production.