How to read AI prediction markets: model names, benchmarks and the fine print
New AI models land almost every week. Here is how Naugur AI markets are worded, and why the rules matter more than the headline.

Keeping up with AI releases is a full time job. Google unveiled Gemini 4 at the end of September and has rolled out new versions since. Anthropic shipped Claude Fable 5.2 in November. OpenAI has released several GPT-6 variants. Each launch brings new names, new benchmark claims and a lot of noise.
Prediction markets cut through some of that noise, but only if you read them the right way.
"Will OpenAI release GPT-7 before July 1, 2027?" resolves on a model actually called GPT-7. A big new model with a different name does not count, however good it is. Companies often add a new suffix instead of a new number, so naming markets usually trade lower than "will there be a better model" would.
On Naugur, a model counts as released when the public can use it in the company's app or API. Research papers, demos and invite only previews do not count. Every market spells this out in its rules.
Leaderboards like LMArena move week to week. A market on who is number 1 on a specific date is really a bet on the release calendar around that date, not on who is "best" overall.
"Will AGI arrive by 2030?" is a fun question with no agreed definition. That is why those markets name a specific test or source in their rules. Read the rules before you trade, they decide the outcome, not the headline.
All AI markets live in the Tech hub. Each one shows its rules, its close date and the sources used to settle it.
Naugur News is written by the Naugur team for information only. It is not financial advice. Market prices move with every trade and can differ from the numbers in this story.

After SpaceX's record listing in June, all eyes are on Anthropic's reported debut. OpenAI looks set to wait until 2027.
AI, space, gadgets, Big Tech and science now live together in one Tech hub. Here is how those markets are written, priced and settled.