Arlo, Lucian, Orion and Curzon.
Four Artemii models, each configured for a different kind of work. Pick one per chat and change your mind at any time.
The four models
Arlo 3.2
The default.
Our fastest, friendliest model. It answers straight away with light thinking, which suits chat, short questions, rewrites and everyday help.
Best for
- Quick questions and chat
- Short drafts and rewrites
- Simple look-ups and summaries
Lucian 3.5
For hard problems.
Our deepest thinker. It takes longer to answer, but works through difficult maths, logic, planning and multi-step problems with more care.
Best for
- Hard problems and puzzles
- Maths, logic and planning
- Complex multi-step reasoning
Orion 3.0
For research and analysis.
Our research and analysis model. It is built for analysing the sources, documents and data you give it, structuring findings and keeping track of what is known and what is not. With web search switched on, it can also search the web, read pages and run deep research reports.
Best for
- Research and open-source intelligence (OSINT)
- Analysing documents, data and sources you provide
- Briefings, timelines and reports
Curzon 3.1
For design and coding.
Our design and coding model. It builds web pages, apps, scripts and interface designs, and fixes and explains code.
Best for
- Writing and fixing code
- Websites, apps and interface designs
- Explaining and reviewing code
Side by side.
Depth is how much thinking goes into an answer. Speed is how quickly it starts and finishes.
| Arlo 3.2 | Lucian 3.5 | Orion 3.0 | Curzon 3.1 | |
|---|---|---|---|---|
| In a word | Quick answers and conversation | Hard problem solving | Research, OSINT and data analysis | Design and coding |
| Depth | ||||
| Speed | ||||
| Pace | Fastest, light thinking | Slower, thinks longest | Slower, thinks longest | Balanced |
| Best for | Quick questions and chat Short drafts and rewrites Simple look-ups and summaries | Hard problems and puzzles Maths, logic and planning Complex multi-step reasoning | Research and open-source intelligence (OSINT) Analysing documents, data and sources you provide Briefings, timelines and reports | Writing and fixing code Websites, apps and interface designs Explaining and reviewing code |
| Choose it when | Most of the time. Start here. | When the answer matters more than the wait. | When you have material to analyse and want it structured. | When you are building or fixing something. |
All four share the same Artemii instructions, tools, safety settings and artifact support. They differ in what they are tuned for, and in how much the AI thinks before it answers.
Pick by the job.
Write a quick reply or answer a question
Arlo. Fast, friendly and light on your allowance.
Solve a hard maths, logic or planning problem
Lucian. Slower on purpose. It works through the steps with more care.
Analyse a pile of documents, data or sources
Orion. Structures findings, and keeps track of what is known and what is not.
Build a page, fix a bug or explain code
Curzon. Writes and reviews code, web pages and interface designs.
How the models use tools.
Web search and deep research need a search provider to be switched on for your account, so they are available on request. Everything else here works today.
| Arlo | Lucian | Orion | Curzon | |
|---|---|---|---|---|
| Web search and reading pages | On every message when web search is available | When you switch on "Search the web" | When you switch on "Search the web" | When you switch on "Search the web" |
| Deep research | Offered when a question looks like research | When you ask for it | When you ask for it | When you ask for it |
| Searching your files | When you have files | When you have files | When you have files | When you have files |
Every model can use your connected services and Mayfair app tools, ask you before changing anything, and make artifacts. Your own settings can switch web tools off.
Not sure? Start with Arlo.
It is the default for a reason. If an answer feels thin, switch the chat to Lucian. For research and analysis of your own material choose Orion, and for design and code choose Curzon.
Where Mayfair's models come from
The Artemii models are built on open-weight AI models published under open licences, which Mayfair adapts and configures. The full notices are on the open-source licences and model notices page.
Start with Artemii
Sign in with your Mayfair ID and ask your first question.