Yupp
Yupp was a crowdsourced AI model comparison platform that paid users to rank answers from competing models and sold the preference data to AI labs.
RIP Yupp
Shut down · March 2026
Yupp let anyone compare answers from competing AI models side by side and earn credits for their verdicts, selling the aggregated preference data back to AI labs. It signed up 1.3 million users and raised $33M led by a16z crypto's Chris Dixon, with Jeff Dean, Biz Stone, Evan Sharp and Aravind Srinivas among the backers. It closed on 31 March 2026, under a year after that round. The labs had moved to expert-sourced data and agentic systems; broad consumer preference signal stopped being the scarce input.
Last known pricing: Free to use; users earned credits for submitting model comparisons
Is Yupp shut down?
Yes. Yupp shut down on March 31, 2026 and is no longer available. Its record is preserved in the ToolDirectory.AI graveyard, our hand-reviewed registry of AI tools that shut down or were acquired.
What to use instead
Okareo
Simulation testing for voice and text AI agents: synthetic users, audio-condition evals, CI/CD gating.
Free Trial
4.84
234
Coval
Simulation and observability platform that tests, monitors, and evaluates AI voice and chat agents.
Free Trial
4.84
262
Judgment Labs
Judgment Labs is a continuous-improvement stack for AI agents — monitoring, failure analysis, and pre-deploy testing.
Paid - Inquire
4.83
245
Overview
What happened to Yupp
Yupp shut down on 31 March 2026, less than a year after raising $33 million. Every employee was let go and the platform went offline.
The product was a crowdsourced AI model comparison site. You asked a question, Yupp answered it with two competing models side by side, and you picked the better response. Those verdicts accumulated into a preference dataset that Yupp sold to AI labs, and users earned credits for contributing. It was a clean two-sided idea: free model access for consumers, training signal for the labs.
Key Features
- Side-by-side answer comparison across competing frontier and open models
- Credits earned for each comparison submitted, redeemable in-platform
- Aggregated human preference data licensed to AI labs
- Public leaderboards ranking models by real user verdicts
- 1.3 million registered users at peak
Why it shut down
CEO Pankaj Gupta wrote that the model capability landscape had changed dramatically in a single year and would keep changing, and that "the future is not just models but agentic systems." The demand Yupp was built to serve moved on. Labs increasingly wanted expert-sourced, high-touch evaluation data of the kind Scale AI and Mercor supply, not broad consumer preference votes — and as attention shifted to agent behaviour, ranking two chat answers stopped being the measurement that mattered.
The funding was never the problem. The round was led by Chris Dixon at a16z crypto and included Google DeepMind chief scientist Jeff Dean, Twitter co-founder Biz Stone, Pinterest co-founder Evan Sharp and Perplexity CEO Aravind Srinivas. Yupp had capital, distribution and 1.3 million users, and still could not find a durable buyer for what it produced.
What to use instead
For comparing model outputs yourself, LMArena runs the public head-to-head leaderboard Yupp was closest to. For teams that need evaluation inside their own stack, the LLM observability and eval tools in our directory cover offline scoring, regression testing and human review workflows.