LMArena AI: AI Model Evaluation Platform 124

LMArena ai, formerly known as lmsys.org, is an AI model evaluation platform focused on crowdsourced AI benchmarks, built by the SkyLab and LMSYS research teams at UC Berkeley. Similar to platforms like V0 or Bolt, the difference is that after you enter a requirement, two models will provide code and render the frontend page so you can rate them. Users can chat with AI for free on this platform and vote to compare and test different AI chat bots.

LMArena AI Features:

• Blind Testing Mode: Users can ask questions to two anonymous AI models, then choose the best response, ensuring fair evaluation.

• Anonymous Battles: Users can interact with multiple anonymous AI chatbots on the platform, ask questions, and receive answers from different bots. This approach allows users to make comparisons without knowing the identity of the models, thereby reducing bias.

• Voting System: Users can vote on different AI responses, helping the platform collect data to evaluate the performance of each model. This crowdsourced approach makes the evaluation results more objective and reliable.

• Style Control: Evaluate a model's ability to follow user instructions and generate content in specific styles.

• Leaderboard: LMArena AI provides a real-time updated leaderboard showing the performance of different AI models. Users can see which models perform best at specific tasks, helping them choose the appropriate tools or services.

• WebDev Arena: The platform has also expanded a feature called WebDev Arena, where users can enter a requirement and the system will generate two different frontend pages for users to rate. This provides developers with an opportunity to test and compare different designs.

LMArena AI also supports multimodal features, allowing users to upload images on their first question to unlock multimodal battles. This means users can not only interact with text models but also communicate with AI through images, enhancing the platform's interactivity and practicality. In addition, LMArena AI has collected over 1 million user votes to calculate Elo leaderboards for more than 100 models, and users can see who the current LLM champion is.

How to Use LMArena AI:

• Visit the Website: Open the official website at lmarena.ai.

• Choose a Feature: Select anonymous battles or access WebDev Arena based on your needs.

• Ask a Question or Enter a Requirement: Type your question or requirement into the corresponding input box.

• View Results: The system will return responses from multiple AIs, and users can compare and vote on these responses.

• Participate in Voting: After the battle ends, users can rate each response to help the platform improve model evaluation.

Our domestic large model DeepSeek-R1 ranks third on LM Arena's overall leaderboard, performing particularly well in technical areas such as "Hard Prompts," "Coding," and "Math."