ITG NowHome › AI glossary › Elo rating
What is an Elo rating?
An Elo rating is a score that goes up when you win and down when you lose, based on who you beat.
- Category Benchmarks
- Also called Elo score
- Updated 20 Oct 2026
Elo rating, explained simply
The Elo system began in chess. Each player has a number. Beating a strong player gives more points than beating a weak one.
AI sites use it for head to head votes. When people pick one model's answer over another, the winner gains points.
This matters because the number only makes sense against other models. A gap of a few points is small and may not be meaningful.
For example, a model with a higher Elo has won more of its matchups. It doesn't mean it wins every time.
On ITG Now, the AI models page shows Elo style scores from public voting sites, with a link to each source.
Related terms
LMArenaLeaderboardBenchmarkIndex pointsAI model
Questions people ask about Elo rating
Is a higher Elo rating better?
Yes. A higher rating means the model won more matchups against others.
Does Elo mean a model always wins?
No. It shows the chance of winning, not a guarantee. Close scores are close to a coin flip.
Elo vs benchmark score: what is the difference?
Elo compares models against each other through votes. A benchmark score comes from a fixed test.