- Tool type
- Model comparison and evaluation platform
- Made for
- Users comparing AI models; developers and researchers evaluating AI
- Runs on
- Web browser at arena.ai
From @arena
Recent important posts · features
- New feature
Arena Agent Mode now supports connectors like GitHub
“Arena update: Connectors are now in Agent Mode! By bringing real-world tools directly into Agent Mode, users can take on even more complex, real-world tasks from start to finish. Our global community of users can now connect tools like Vercel, Linear, GitHub, Supabase, Notion,”
@arena · Oct 27View
What it does
Arena is a web platform for comparing AI models on real-world prompts. You type a prompt, and in the default Battle mode two anonymous models respond side by side. You vote for the response you prefer, and those votes feed public leaderboards for text, search, vision, and document tasks. Requests for other formats, such as video, are routed to Arena-specific experiences when the prompt makes that clear.
What sets Arena apart is its crowdsourced comparison. Model identities are revealed only after you vote, and follow-up prompts keep using the same pair until a new vote is cast. You can also pick a specific model with the @ selector, or compare two chosen models head to head. Chat mode offers a more traditional AI chat experience, while Work mode is an experimental agentic option for complex, ongoing tasks, with Connectors that can bring in tools such as Gmail and Google Drive.
Arena suits people who want to see how models handle everyday prompts or who evaluate AI for their own work. It is used through the arena.ai website. Leaderboard rankings reflect human preferences from voters rather than a fixed benchmark, and experimental features such as Work mode may not be available to every user.
What it can do
- Anonymous head-to-head comparison of two AI models in Battle mode, with a vote after each response
- Specific model selection with the @ selector, filterable by modality
- Public leaderboards for text, search, vision, and document models
- Chat mode for a traditional AI chat experience
- Experimental Work mode (Agent Mode) for complex, ongoing agentic tasks
- Arena-specific modes such as Video Arena, triggered by the prompt
Works with
- Gmail
- Google Drive
- Google Calendar
- Vercel
- Supabase
- Linear
Before you choose
- Votes in Battle mode feed public leaderboards, so rankings reflect human preference rather than a fixed benchmark.
- Work mode and the simplified chat-bar UI are experiments and may not be available to all users.
- Model availability and the specific models offered can change over time.
Listed under
Get started
Type a prompt into the 'Ask me anything' text box.
- Go to arena.ai and type a prompt into the 'Ask me anything' text box.
- Press Enter or click the submit button to see two anonymous model responses.
- Vote for the response you prefer; both models are then revealed.
Plans and prices
Confirm the applicable plan, usage charges and commercial terms with the provider.
We couldn’t find published prices for Arena. See their pricing page ↗
Sources for this overview
Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.
- About Us - Arena.ai ↗arena.ai
- Arena Experiments: Simplified UI ↗help.arena.ai
- How to Compare AI Models in Battle Mode ↗help.arena.ai
- Chat Arena Leaderboard: Best AI Chat Models & LLM Rankings ↗arena.ai
- How to Select Specific Models in Arena ↗help.arena.ai
- LMArena is now Arena - Arena.ai ↗arena.ai
Ratings and reviews
No ratings for Arena yet.
What’s changed
Free accountSignals
Free accountData and security
Free accountCompare with similar tools
Free accountWhere this comes from
How we research- Company website
- arena.ai: product, pricing and policy pages
- YC directory mirror
- Batch and company name
- Social accounts
- Accounts the company links, and its posts on X
- Human review
- Checked by a person
- knowledge_base
Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.