Yupp is shutting down less than a year after launch, despite raising $33 mn and attracting backing from one of Silicon Valley’s best-known AI investor circles.
Yupp raised its $33 mn seed round in 2024, led by Chris Dixon of a16z crypto. For the period, it was a huge seed round.
The startup also said it brought in more than 45 angel investors and smaller backers, including Google DeepMind chief scientist Jeff Dean, Twitter co-founder Biz Stone, Pinterest co-founder Evan Sharp, and Perplexity chief executive Aravind Srinivas.
The startup built a crowdsourced service for comparing AI models. Users could test results from a pool of 800 models for free, including leading systems from OpenAI, Google, and Anthropic.
Yupp returned multiple answers to a single prompt, whether text or images, then asked users to rate which model worked best and explain why.
The business thesis was simple enough. Collect anonymised data on what people want from AI, then sell that signal to model developers.
Yupp said it reached 1.3 mn users and gathered millions of preference signals each month. It also built a leaderboard and said several AI labs had already signed on as customers.
We launched last June to build a two-sided marketplace for AI model evaluation, giving users access to multiple models and using that real-world usage to help AI labs evaluate them. We signed up more than 1.3 mn users, and several labs as paying customers; but we didn’t reach a strong enough product–market fit.
Pankaj Gupta, Co-founder, Yupp
Meanwhile, the landscape has shifted quickly. Models are getting much more capable, and user workflows are moving toward agentic systems: models connected to tools, memory, and external services, rather than standalone chatbots. In that world, crowdsourced model evaluation at the chatbot layer becomes increasingly less critical.
That still wasn’t enough. Co-founders Pankaj Gupta and Gilad Mishne said the company failed to reach strong enough product-market fit to keep going.
We’re proud of what we built together here. One of the things that meant the most to us was being able to give our community free access to the world’s top AI models.
Pankaj Gupta, Co-founder, Yupp
“You used that access to learn, to build and share wonderful things, to shape AI with your feedback, and to grow in your careers; and then you brought those skills home to your friends, family, and communities. That meant everything to us”, Gupta says.
Part of the problem, they said, was the speed of change in the model market. AI systems improved too quickly over the past few months, making the original positioning harder to hold.
Labs are still paying heavily for feedback, though the model has shifted.
Companies such as Scale AI and Mercor have pushed a different approach, hiring specialist experts, including PhDs, and placing them inside reinforcement learning workflows.
At the same time, the industry is already looking further ahead. More builders now focus on a future where AI systems are built for other AI systems, and increasingly used by them too. Consumer feedback still has value, though the direction of travel is changing.
Gupta said the AI capability landscape has changed dramatically over the last year and will keep changing fast. In his view, the future is no longer only about models. It is about agentic systems.
We know this is disappointing news. We’re grateful to you and are sorry for this outcome. We have no doubt you’ll all go on to do great things with the knowledge you’ve built here.
Gupta said some Yupp employees are moving to a well-known AI company, while others are looking for their next role. Big round, strong network, smart pitch. “Finally, we’d like to thank our customers, investors and our incredibly talented team to have made this journey possible”.
