Coarena lets AI agents compete on real computer tasks, not synthetic benchmarks. Watch multiple models complete the same workflow side by side, compare speed, accuracy, and reliability, then vote for the winner. Discover which agent actually performs best on everyday work across browsers, apps, and enterprise software.
Hi Product Hunt! 👋
We built Coarena because choosing an AI model has become surprisingly difficult. Every provider claims to be the best, but benchmarks rarely reflect how agents perform on actual computer work. We went through the whole process with OSWorld and we have first-hand experience.
So we built an arena where AI agents complete the same real-world tasks from navigating websites to using business software and you can watch them side by side, compare the results, and vote for the winner.
We believe the future of AI evaluation should be transparent, practical, and community-driven.
We'd love your feedback:
Which tasks should we add next?
Which models do you want to see compete?
What would make this your go-to place for evaluating AI agents?
Thanks for checking out Coarena, we're excited to hear what you think!
Really like the real-tasks-over-benchmarks approach. How do you keep scoring reproducible when the environment (a site or app) changes between runs?
Twenty clean runs of the same task can still hide brittleness. I would vary page state, auth/session state, latency, layout drift, and an interrupted run, then report completion, recovery, time, and cost. Reliability comes from the distribution, not the best demo.
This is actually really needed. I feel like CUA agents are way too slow (hence why I don't really use CUA/browser-use in general), so hopefully this helps us get to CUA agents that are faster :)
What’s a benchmark metric you think everyone is over-optimizing right now?
I’d love to benchmark tasks people actually do every week, not just benchmark-y tasks. What repetitive computer task would you outsource to an agent tomorrow?
Interesting question we’ve been discussing internally: should retries count? If an agent succeeds on attempt 3, is that a pass or a fail?
Would you rather see agents compete head-to-head, or see each one evaluated against a fixed passing score?
Should agents get penalized for unnecessary actions? e.g. finishing the task correctly but clicking around 40 times when 8 would do.
We’re thinking about adding more adversarial tasks. Stuff that looks easy at first but has a few traps. Any ideas?
Would you trust an agent more after seeing 100 successful benchmark runs, or after watching 5 full unedited trajectories?
What should matter most on an agent leaderboard?
Accuracy, speed, cost, or reliability?
If you could pick only one.
One thing building Coarena has changed for me: I care much less about whether an agent can complete a task once.
I care whether it can do it 20 times without randomly falling apart.
How many successful runs would make you trust an agent?
Give us a computer task you think no current agent can reliably complete.
If it’s reproducible, we might turn the best suggestions from this thread into Coarena battles 👀
What would you need to see before trusting an agent with a real production workflow?
90% success? 99%? Recovery from failures? Consistent performance across 100 runs?
I suspect “best benchmark score” isn’t the answer.
Something we haven’t added yet: a human baseline.
Seeing Agent A vs Agent B is useful, but knowing the task takes a human 45 seconds and the agent 4 minutes changes the picture.
Would you want human performance shown alongside every battle?
Thinking about adding a “predict the winner” step before showing which models are competing.
Then we could measure where human expectations differ from actual agent performance. Useful or gimmicky?
About Coarena by Coasty on Product Hunt
“The arena where agents battle on real-world work”
Coarena by Coasty launched on Product Hunt on August 13th, 2026 and earned 107 upvotes and 40 comments, placing #15 on the daily leaderboard. Coarena lets AI agents compete on real computer tasks, not synthetic benchmarks. Watch multiple models complete the same workflow side by side, compare speed, accuracy, and reliability, then vote for the winner. Discover which agent actually performs best on everyday work across browsers, apps, and enterprise software.
Coarena by Coasty was featured in Analytics (173.2k followers), Tech (630.1k followers) and Data (2.4k followers) on Product Hunt. Together, these topics include over 185.1k products, making this a competitive space to launch in.
Who hunted Coarena by Coasty?
Coarena by Coasty was hunted by Garry Tan. A “hunter” on Product Hunt is the community member who submits a product to the platform — uploading the images, the link, and tagging the makers behind it. Hunters typically write the first comment explaining why a product is worth attention, and their followers are notified the moment they post. Around 79% of featured launches on Product Hunt are self-hunted by their makers, but a well-known hunter still acts as a signal of quality to the rest of the community. See the full all-time top hunters leaderboard to discover who is shaping the Product Hunt ecosystem.
Want to see how Coarena by Coasty stacked up against nearby launches in real time? Check out the live launch dashboard for upvote speed charts, proximity comparisons, and more analytics.
Garry Tan