Frontier model
A frontier model is one of the most capable general-purpose AI models available at a given time, usually from a well-funded lab. The term is relative and also appears in AI safety policy.
Also called: frontier lab, frontier AI
By Divya Sharma · Checked Nov 15, 2026 · How we review
"Frontier" describes position, not a technical feature. A model is frontier if it is at or near the top on capability when it ships. A model that was frontier two years ago usually is not now. The labs that build these models are often called frontier labs.
Two uses of the word
In everyday talk, frontier model means the strongest large language models on hard benchmarks and leaderboards. On this site, the rankings show which models are at the top right now.
In policy writing, the term is narrower. A 2023 paper on frontier AI regulation defined frontier AI as highly capable foundation models that could have capabilities dangerous enough to threaten public safety. Laws and voluntary safety frameworks have since used their own versions of the idea. These definitions differ, so check which one a given rule uses; this page does not give legal advice.
How it differs from open-weight
Frontier and open-weight are separate questions. Most frontier models are only available through an API, but some open-weight models get close to the frontier. "Frontier" says how capable a model is; "open-weight" says whether you can download it.
Why it matters when building
Frontier models usually cost the most per token and can be slower. Many tasks do not need one. A common pattern is to send hard requests to a frontier model and routine ones to a smaller model, which is what LLM routing does.
Related terms
- Large language model
- A large language model (LLM) is a neural network with billions of parameters, trained on large amounts of text to predict the next token, which lets it write, answer questions, summarise and follow instructions.
- Open-weight model
- An open-weight model is a model whose trained weights are published for anyone to download and run, often without the training code or full training data, so it is not necessarily open source.
- AI benchmark
- An AI benchmark is a fixed, public set of tasks with an agreed scoring method, used to compare models on the same test, for example coding issues solved, exam questions answered or retrieval quality across many datasets.
- LLM routing
- LLM routing is the practice of sending each request to the language model best suited to it, often a cheap model for easy prompts and a stronger one for hard prompts, to balance cost, speed and quality.
- Small language model
- A small language model (SLM) is a language model with relatively few parameters, usually a few billion or fewer, built to run on a laptop, phone or single GPU at lower cost than a large model.