Model Rankings

Model rankings are used to help compare the relative performance of different models and should not be used alone as the basis for model selection. Data comes from Arena AI, and is automatically updated daily by GitHub Actions.

When choosing a model, you should also consider task type, context length, multimodality and tool-calling capabilities, speed, price, regional availability, and data policies. Rankings and prices change dynamically; please refer to the update time marked on the page and the official information from the model provider.

Ranking directory

• Agent rankings: This ranking evaluates models' overall performance on multi-turn tool-calling / agent tasks, covering metrics such as net improvement rate, confirmation success rate, thumbs-up/down ratio, controllability, Bash recovery rate, and tool hallucination rate.
• Text rankings: This ranking is Arena AI's core text battle ranking, estimating models' relative strength based on human blind-vote battles.
• Search rankings: This ranking evaluates models' performance on web-connected search / information retrieval tasks.
• Vision rankings: This ranking evaluates the performance of multimodal models on image understanding tasks.
• Code / Web development rankings: This ranking evaluates models' practical performance in Web front-end development (HTML/CSS/JS) tasks.
• Text-to-image rankings: This ranking evaluates the ability of text-to-image models to generate images from text prompts.

---

💡 Get help and submit feedback

If you encounter any questions, bugs, or have suggestions for feature improvements during configuration or use, please refer to Feedback and suggestions for the official channels provided.