Compare AI models such as ChatGPT, Claude, Gemini, Qwen and DeepSeek by vote score, price and context length. The data updates itself. Free.
Loading the leaderboard...
How to read the leaderboard
- Ranking score: Calculated with an Elo-style system from millions of anonymous side-by-side comparisons of two answers. The higher the score, the more often people chose the model for the task type you are viewing.
- Confidence interval: The small line under each score shows the range the score could plausibly fall in.
- Votes: The more votes, the more reliable the score. New models usually have fewer votes, so their confidence interval is wider.
- Open and closed: Open models can be downloaded and run yourself (under each model's license), while closed models are only available through the provider's service. Hover over the badge to see the license name.
- Price and context: In the OpenRouter tab, prices are per 1 million tokens (roughly 750 thousand words). Context is how much text the model can read in one go.
Frequently asked questions
The server automatically pulls fresh data from the sources whenever its copy is more than about 25 minutes old, and your browser re-checks every 10 minutes while the page is open. Prices and the model list on OpenRouter change constantly, so they update almost instantly. LMArena publishes its score tables in batches, and the publication date is shown right on the leaderboard.
Not necessarily. The score reflects what most users prefer, while your work may need speed, cost, a specific language or security. Filter by the right criteria, then try 2 to 3 models on your real tasks.
Ranking scores come from the public LMArena dataset. Prices, context length and release dates come from OpenRouter’s public model list. The two sources measure different things, so model names may not match exactly.
Logos are matched by company and model family. Smaller companies without a logo in the collection show initials instead. Logos and brand names belong to their respective owners and are used for identification only.
It depends on the task: writing content, customer support, coding or data analysis will each give different results. If you would like advice specific to your business, please contact Nam.
