What is lm arena?
- User
What is lm arena?
- Assistant
LM Arena (Language Model Arena) is a platform created by researchers to evaluate and compare different large language models through human preferences. It works like a competitive ranking system where users can interact with two anonymous AI models side-by-side, ask them the same question, and then vote for which response they prefer.
The platform uses an Elo rating system (similar to chess rankings) to create a leaderboard based on thousands of these human preference votes. This crowdsourced approach helps provide a more realistic assessment of model performance compared to traditional benchmarks, since it captures what actual users find helpful in real conversations.
The arena includes various commercial and open-source models from companies like Anthropic, OpenAI, Google, Meta, and others. It's become an influential benchmark in the AI community because it reflects real-world usage patterns and human preferences rather than just technical metrics.
The platform is maintained by researchers primarily from UC Berkeley and has been useful for tracking the progress of different language models over time as new versions are released.
Other models, same or similar prompt
Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.
Identical first prompt
- llama-4-maverick-03-26-experimentalCompare side by side →
- claude-opus-4-20250514Compare side by side →
- gemini-2.5-flashCompare side by side →
- grok-4-0709Compare side by side →
- chatgpt-4o-latest-20250326Compare side by side →
- qwen-max-2025-01-25Compare side by side →
- command-a-03-2025Compare side by side →
- claude-3-7-sonnet-20250219-thinking-32kCompare side by side →
- mistral-medium-2505Compare side by side →