Best AI for hard prompts: 417 models ranked by blind votes

Blind votes

LMArena puts the same prompt to two anonymous models and asks the person who wrote it to pick the better answer. The votes place every model on one scale, where a gap of 100 points means the higher model wins about 64 times in 100.

Style control

The scores discount answer length and formatting, so a model does not climb by writing more or by adding headings and lists.

Shared ranks

A model’s rank is one plus the number of models whose whole 95% interval sits above its own. Models the votes cannot separate share a rank, which is why several can be first.

Settings

A model can appear more than once, at different reasoning settings: the line under its name gives the entry as LMArena lists it. Entries with web search switched on are a different product and carry no price.

Prices

The list price per million tokens from the maker’s own pricing page, read on 15 October 2026. It is a price per token, not per answer: a higher reasoning setting spends more tokens on the same question.

Source: LMArena, leaderboard dataset (revision 9041150e, published 20 October 2026), licensed under CC BY 4.0. We change no score: we add the prices and the links to our model pages.