Early-stage project | Chance AI secures millions of dollars in investment from Meitu and others, with user count reaching 200,000.

TL;DR

• Chance AI, a Visual Agent startup founded in 2025, has raised several million dollars in an angel round led by Meitu, with follow-on investment from NYX Ventures and an Alibaba-affiliated fund.
• The company's core product positions itself as the world's first AI application using the camera as the primary interaction entry point, bypassing traditional text input boxes.
• Their Visual Agent achieves an 86.07% accuracy rate on the MMMU-Pro multimodal reasoning benchmark, surpassing the human baseline of 85.4%.
• The platform has accumulated approximately 200,000 users, with 40% located in the North American market, covering over 35 countries.
• User retention is strong, with a 30-day revisit rate of 49.2%, and the app has reached the top spot on Product Hunt twice.

Why It Matters

This project signals a shift from text-centric AI interactions to vision-centric "agent" workflows, where AI interprets intent and suggests actions based on visual context rather than just recognizing objects. For practitioners and investors, it highlights a new category of consumer AI that integrates deeply into personal daily life decisions (fashion, image, social) and targets the "Visual Native" demographic, offering a strategic counterpoint to general-purpose LLMs.

Key Data

• Benchmark Performance: Chance AI's Visual Agent scored 86.07% on the MMMU-Pro multimodal reasoning benchmark.
• Human Baseline Comparison: The system's accuracy exceeds the established human baseline of 85.4% on the same benchmark.
• User Traction: The product has secured approximately 200,000 cumulative users.
• Retention Rate: The 30-day revisit rate stands at 49.2%.
• Geographic Split: Roughly 40% of the user base is from North America, with the product available in over 35 countries.

Technical Details

• Interaction Paradigm: The system utilizes a "see — understand intent — invoke Agent — complete action" logic, distinguishing it from standard "photo — recognition — result" pipelines by focusing on the user's underlying goal and subsequent actions.
• Multimodal Reasoning: The core technology is optimized for multimodal inference, demonstrated by its top ranking on the MMMU-Pro benchmark, indicating advanced capabilities in integrating visual and textual data to solve complex problems.
• Personalized Visual Memory: As usage increases, the system constructs a personal "visual memory" for each user, tracking style preferences, wardrobe composition, and social image to tailor future responses and suggestions.
• Founding Context: Founded by Zeng Xi, who holds a PhD in Cognitive Science and Contemporary Art from the University of Barcelona and previously worked on AI products at ByteDance and hardware companies like OnePlus/OPPO, blending academic cognitive science with industrial hardware and software experience.

Industry Insight

• Shift to Vision-First Interfaces: The success of Chance AI suggests that for specific consumer verticals (fashion, personal style), camera-based interaction is superior to text-based prompts, as it aligns with how "Visual Native" users naturally process information.
• Community-Based Monetization: The long-term strategy moves from a pure tool to an "AI-native lifestyle community," implying that future monetization may rely on social sharing and community engagement (subscriptions for sharing tools) rather than just API usage or basic feature unlocks.
• Target Demographic Focus: By aggressively targeting North American university students, particularly young women, the company is establishing a high-LTV (Life Time Value) cohort early, using offline campus activations to drive seed user acquisition, a model that could be replicated by other niche AI startups seeking high-engagement niches.

ze through three channels: subscriptions for advanced features, hardware licensing, and cautious advertising recommendations. However, their current priority is user habit formation before aggressively pushing monetization.