a16z Leads $33M Seed Round: How Yupp is Leveraging Blockchain and Incentives to Reshape AI Evaluation Models?

Original Title: "a16z Leads $33M Seed Round: How Yupp is Reshaping AI Evaluation Models with Blockchain and Incentives?"

Original Author: ShenZhen, PANews

As AI applications permeate multiple industries, accurately evaluating model performance and building user trust have become pressing issues. Traditional evaluation methods often rely on centralized systems, which struggle to cover diverse use cases and fail to capture real user preferences. At the same time, the issue of "hallucinations" in AI models frequently arises, leaving users trapped in echo chambers when making choices.

Against this backdrop, Yupp, an emerging platform, is leveraging its unique crowdsourcing model and incentive mechanisms to redefine the way AI models are discovered, compared, and utilized. The platform aims to introduce a paradigm shift in AI evaluation. This article will provide an in-depth analysis of Yupp’s core mechanism, technical highlights, team background, and its potential impact on the AI ecosystem.

Team Background and Funding: Backed by Big Tech Expertise

Yupp is focused on addressing the long-standing challenges in AI evaluation. It is building a trustless AI feedback marketplace where diverse user feedback can flow freely, supported by blockchain technology and crypto-economic incentives, to create a scalable, fair, and transparent model evaluation layer. By incentivizing the distribution of high-quality human-labeled data, Yupp captures real-world user needs and preferences across various scenarios in a timely manner. This, in turn, helps AI developers iteratively optimize their model performance.

The project was founded in June 2024 by Pankaj Gupta (Co-founder and CEO) and Gilad Mishne (Co-founder and Head of AI), with Jimmy Lin (Professor at the University of Waterloo) serving as Chief Scientist. The trio initially worked together at Twitter in 2010, where they developed and fine-tuned large-scale recommendation and search systems. They later gained extensive experience at Google and Coinbase.

Thanks to its vision of decentralization and creating transparency in data value, Yupp addresses AI companies’ twin concerns around trustworthy evaluation and user participation. Combined with the impressive credentials of its core team, the startup has received high praise from prominent figures in the tech industry and leading venture capital firms.

Last week, Yupp announced the completion of a $33 million seed round led by a16z partner Chris Dixon. Other investors include Google Chief Scientist Jeff Dean, Twitter Co-founder Biz Stone, Pinterest Co-founder Evan Sharp, Perplexity CEO Aravind Srinivas, and renowned academics and executives such as Stanford University’s Dan Boneh, Chris Ré, Nick McKeown, and Balaji Prabhakar. The round also saw participation from 45 notable angel investors and executives, as well as Coinbase Ventures.

Core Features and User Experience: Building the "AI Assembly"

As a centralized AI evaluation platform, Yupp adheres to the philosophy of "Every AI for Everyone," enabling users to easily discover, compare, and utilize the latest AI models. Unlike traditional single-response setups, Yupp simultaneously returns answers from two (or even more) models for every prompt, creating an "AI Assembly." This design not only caters to users' need for diverse options but also effectively identifies potential "hallucinations" in models, helping users make more informed decisions through comparisons. As Yupp CEO Pankaj Gupta puts it, side-by-side output is especially beneficial for users concerned about generative errors since it allows for cross-validation of the results.

The platform now supports over 500 AI models across text and image generation, including renowned models like ChatGPT, Claude, Gemini, DeepSeek, Grok, Llama, and many emerging models. To further enhance the experience, Yupp has introduced the "QuickTake" feature, which condenses lengthy responses into a concise tweet-like summary.

Moreover, Yupp places a high priority on user privacy: all chat records are private by default unless explicitly shared by the user; even when sharing, no personal information is disclosed. Users retain full control over what, how, and with whom their content is shared.

Economic Model and Incentive Mechanism: Valuing Data Labor

Yupp integrates free usage with user feedback through its "Yupp Credits" system, which measures model interactions. New users receive 5,000 credits upon registration, and they can earn additional credits by rating model responses, selecting preferences, and providing explanations. The higher the quality of the feedback, the greater the rewards, ensuring users can continue to access high-end models like Claude Opus 4 or OpenAI o3 for free. The platform guarantees that credits only increase and all models are currently free to explore.

After each query, users receive two model responses and can win a "digital scratch card" through their feedback, rewarding them with 0–250 Yupp Credits. Every 1,000 credits equate to $1, with a maximum daily withdrawal of $10 and a monthly cap of $50. Credits can be redeemed in over 20 currencies, including USD and EUR, through partners such as Stripe, PayPal, and Coinbase. Additionally, the platform integrates Base Ethereum L2 and Solana stablecoins, offering global users instant, fee-free rewards.

As Pankaj Gupta mentioned, high-quality user-generated feedback holds far greater value for AI companies in fine-tuning models and enhancing reinforcement learning than the rewards themselves. While monthly user earnings might equate to just a few cups of coffee, these paid annotation datasets are critical for AI iteration.

To encourage more participation, Yupp has introduced referral rewards: the referrer receives 5,000 points, while the referred user gets 1,000 points. Currently, new users signing up can earn 5,000 points, with referred users receiving an additional 2,500 points.

Yupp VIBE Score: A New Paradigm in AI Evaluation

Addressing issues with current leaderboards such as lack of transparency, fairness concerns, and unequal access to evaluation data, Yupp has launched a beta AI leaderboard and the “Yupp VIBE (Vibe Intelligence Benchmark) Score” system. This system aggregates user preference data from natural interactions worldwide to deliver robust and reliable evaluation results.

Yupp's evaluation principles include:

· Robustness: Ensuring representativeness (covering diverse scenarios), authenticity (reflecting user concerns), and cheat-resistance (guarding against malicious behavior);

· Trustworthiness: Upholding fairness (remaining unbiased toward models), transparency (openly disclosing ranking algorithms), and scientific rigor (adhering to evaluation protocols).

The platform not only collects binary preferences but also encourages users to highlight pros and cons of the responses (e.g., "spot-on," "fast response," "great style"). Additionally, user information such as age, education, and profession is utilized for segmentation analysis, providing insights into preference differences across demographic groups.

On a technical front, Yupp is exploring the use of blockchain, cryptographic primitives, and zero-knowledge proofs to ensure the evaluation process is fair, transparent, and verifiable. Furthermore, the platform has partnered with professional AI data providers to calibrate scorers through profile verification and multi-layer quality checks, effectively filtering out malicious data.

Recent leaderboard updates showcase the VIBE scores of models such as GPT-4.5 Preview, Claude Opus 4, and Claude Sonnet 4, alongside metrics like win rate, dislike rate, speed, latency, context window, and cost indicators.

Development Timeline and Future Outlook

Yupp officially launched on June 13, 2025, following a six-month internal testing phase. Since its launch, the product has undergone continuous iterations:

· Multi-modal support: Integrated models such as Dall‑E, Flux, Stable Diffusion, Luma Photon, Google Imagen 4, and supported user-uploaded images/PDFs for Q&A;

· Interaction channel expansion: Added voice input and voice reading functionalities;

· Model updates: Gradually introduced DeepSeek R1/V3, Mistral Small 3, OpenAI o3‑pro, Hermes 3, Amazon Nova Pro v1, Microsoft Phi series, and the "MAX Model" category;

· Real-time information: Routed online query requests to Perplexity and Google Gemini Live, accompanied by hyperlinked citations;

· Payment upgrades: Added U.S. PayPal, Venmo withdrawals, and support for 24 currencies via PayPal;

· Sharing and export: Enabled format-preserving copying, PDF/text/Markdown export, and selective sharing of individual responses or entire conversations on demand;

· Community activities: Hosted events like the “AI Prompt Challenge” with prizes worth tens of thousands of points; introduced personal profile pages and AI-generated conversation names.

Yupp’s mission is to "empower humanity to shape the future of AI." Pankaj Gupta believes that AI’s development requires participation and contributions from everyone. Through multi-perspective AI responses and user feedback, Yupp not only assists users in making better decisions but also provides continual momentum for AI evolution.

It is noteworthy that one of Yupp’s primary competitors is the open AI model evaluation platform LMArena (website: https://lmarena.ai/), which is highly popular among AI professionals. However, the platform is currently in its commercialization exploration phase and does not leverage blockchain technology to provide direct material rewards or points-based incentive mechanisms for user participation.

Overall, Yupp has paved a new path in AI evaluation with its crowdsourced model, incentive mechanisms, and user-preference-driven evaluation system. It not only provides free and diverse AI interaction experiences for users but also transforms user feedback into high-value training data, driving continuous model optimization. Backed by an experienced team and top-tier capital, Yupp is poised to play a pivotal role in the future AI ecosystem, realizing the vision of "AI for all, shaped by all."

However, for the newly launched Yupp, how to continuously ensure data quality under large-scale user participation, resist potential cheating behaviors, and strike a balance between commercialization and user incentives will remain key areas that require ongoing exploration and optimization in its future development.

Original Link

Welcome to join the official BlockBeats community:

Telegram Subscription Group: https://t.me/theblockbeats

Telegram Discussion Group: https://t.me/BlockBeats_App

Official Twitter Account: https://twitter.com/BlockBeatsAsia