Key Takeaways
- • Yupp, a crowdsourced AI model-comparison startup, is shutting down less than a year after launch, according to co-founders Pankaj Gupta and Gilad Mishne, who announced the closure on Tuesday.
- • The company raised a $33 million seed round in 2024 led by a16z crypto's Chris Dixon, plus checks from more than 45 angel investors including Jeff Dean, Biz Stone, Evan Sharp and Aravind Srinivas.
- • Yupp said it signed up 1.3 million users and collected millions of preferences monthly, and counted a few AI labs as customers, yet failed to reach "a strong enough product-market fit."
- • The founders attributed the failure partly to rapid improvements in AI models over recent months, which eroded the value of the feedback Yupp sold.
- • Industry momentum is shifting toward specialist expert feedback (as used by Scale AI and Mercor) and toward agentic AI systems, away from consumer preference data.
Summary
Yupp is closing its business less than a year after launch, co-founders Pankaj Gupta and Gilad Mishne announced on Tuesday, according to TechCrunch. The startup had raised a $33 million seed round in 2024, led by Chris Dixon of a16z crypto and described at the time as a large seed round for its day. The company also raised checks from more than 45 angels and small investors, including Google DeepMind chief scientist Jeff Dean, Twitter co-founder Biz Stone, Pinterest co-founder Evan Sharp and Perplexity CEO Aravind Srinivas.
Yupp offered a crowdsourced AI model-picking service. Consumers could test and compare responses from roughly 800 AI models, including models from OpenAI, Google and Anthropic, for free. Users gave feedback on which outputs worked best and why. The business model rested on generating anonymized data about what people need from AI, which model developers would then pay for. Yupp said it signed up 1.3 million users, collected millions of preferences every month, operated a leaderboard and had a few AI labs as customers.
Despite these metrics, Yupp did not reach a strong enough product-market fit to sustain the business. The founders cited rapid gains in AI model capability over the past few months as a contributing factor. Gupta, Yupp's CEO, wrote on X that "the AI model capability landscape has changed dramatically in the last year alone and will continue to change quickly," adding that "the future is not just models but agentic systems."
The closure also reflects broader shifts in the AI data market. Labs are increasingly paying for feedback from specialist experts, such as PhDs embedded in reinforcement learning loops, a model pioneered by companies including Scale AI and Mercor. Meanwhile, much of Silicon Valley's attention has moved toward AI systems built for use by other AI agents, though model makers still value some consumer feedback in the near term.
Gupta said some Yupp employees are joining a "well-known" AI company, and others are seeking new roles. Yupp did not immediately respond to TechCrunch's request for comment.
Yupp offered a crowdsourced AI model-picking service. Consumers could test and compare responses from roughly 800 AI models, including models from OpenAI, Google and Anthropic, for free. Users gave feedback on which outputs worked best and why. The business model rested on generating anonymized data about what people need from AI, which model developers would then pay for. Yupp said it signed up 1.3 million users, collected millions of preferences every month, operated a leaderboard and had a few AI labs as customers.
Despite these metrics, Yupp did not reach a strong enough product-market fit to sustain the business. The founders cited rapid gains in AI model capability over the past few months as a contributing factor. Gupta, Yupp's CEO, wrote on X that "the AI model capability landscape has changed dramatically in the last year alone and will continue to change quickly," adding that "the future is not just models but agentic systems."
The closure also reflects broader shifts in the AI data market. Labs are increasingly paying for feedback from specialist experts, such as PhDs embedded in reinforcement learning loops, a model pioneered by companies including Scale AI and Mercor. Meanwhile, much of Silicon Valley's attention has moved toward AI systems built for use by other AI agents, though model makers still value some consumer feedback in the near term.
Gupta said some Yupp employees are joining a "well-known" AI company, and others are seeking new roles. Yupp did not immediately respond to TechCrunch's request for comment.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.