Xinchen AI Secures Another Strategic Investment from Outfit7, Launches Self-Developed General-Purpose "Xinchen Large Model" | BlueRun Ventures

On July 9, marking its second anniversary, Xinchen AI (Hangzhou) Technology Co., Ltd. (hereinafter referred to as "Xinchen AI") officially launched its self-developed general-purpose large language model — the "West Lake Model." Zhenzhong Lan, the company's founder and a doctoral supervisor at Westlake University, stated at the model's launch event: "Build a general-purpose large model with both high IQ and EQ, creating AI that companions one billion people."

Recently, Xinchen AI secured another strategic investment from Outfit7 Limited (Talking Tom), with the investment amount exceeding 100 million yuan. The two parties will accelerate the application and commercialization of the large model.

BlueRun Ventures was the lead investor in Xinchen AI's angel round and has continued to increase its stake in subsequent rounds.

Currently, Talking Tom boasts 400 million monthly active users globally. After fully integrating the West Lake Model, the annual model call volume is expected to potentially reach hundreds of billions.

Since its establishment, Xinchen AI has attracted investments from multiple well-known institutions including Talking Tom, BlueRun Ventures, Baidu Venture, Kaitai Capital, Xihu Sci-Tech Innovation Investment, and the Westlake University Education Foundation Sustainable Development Platform.

01

Getting Hands Dirty, Standing Out

Xinchen AI is among the earliest large model companies in China to launch a series of 2C products (AI writing, AI drawing, AI psychological counseling), having accumulated over two million C-end users and served hundreds of B-end clients, including well-known enterprises such as Talking Tom, Zhiyi, Starbucks, Alipay, Zhihu, and Kujiale.

Xinchen AI's ability to stand out among internet giants from the very beginning stems from its top-tier technical team with high talent density.

The team's founder, Zhenzhong Lan, after earning his Ph.D. from Carnegie Mellon University (CMU), worked as an AI scientist at Google.

In 2018, Lan used the computing power of thousands of TPUs to successfully develop "ALBERT," a lightweight version of Google's large model BERT (with over 5,000 academic citations). This research has been applied to products covering hundreds of millions of users, including Google News and Google Assistant, exerting profound influence in the large model field.

Under Lan's leadership, Xinchen AI has assembled a group of top AI talents from companies such as Google, Meta, and Amazon.

More importantly, Xinchen AI is backed by Westlake University, a new type of research university in China. Its abundant talent reserves and powerful computing resources have laid a stable foundation for building the West Lake Model with both high IQ and EQ.

Following this funding round, Xinchen AI and Talking Tom will accelerate cooperation in general-purpose large models, vertical domain models, and terminal software and hardware.

Talking Tom stated that compared to other large models, the West Lake Model features long-term memory, emotional perception, and proactive chatting capabilities, while integrating multiple modalities including text, images, and voice — highly aligned with Talking Tom's scenarios and future plans.

The two parties are accelerating large model deployment, jointly launching a series of Talking Tom products that can listen, see, speak, teach, and companion.

02

Forging a Different Path, Dual Intelligence Online

Currently, most large model development focuses on the model itself while neglecting the human element (the interaction partner). For instance, ChatGPT's highly neutral, knowledge-based, tool-like persona, despite its vast knowledge, generally lacks personality and struggles to form emotional connections with users.

Unlike correct but cold responses, the West Lake Model is human-oriented — more human-like, more attuned to human hearts.

The U.S. large model company Character.ai allows users to chat with AI virtual characters, becoming a unicorn within six months of founding, with monthly visits exceeding 300 million.

Notably, Character.ai's average user interaction duration exceeds 28 minutes, while ChatGPT's is less than five!

Another U.S. large model company, Inflection.ai, recently completed a $1.3 billion funding round. Its chatbot Pi, compared to ChatGPT, possesses stronger empathy and more easily forms emotional connections with people.

The rise of Character.ai and Inflection.ai demonstrates that "high-EQ" large models have become a highly promising and capital-favored track.

Large models with "high EQ" can better interact with humans, more deeply understand and satisfy human needs, thereby making users more willing to engage with AI.

Currently, the vast majority of large models represented by ChatGPT only respond matter-of-factly to user inputs, unable to proactively gather information or deeply understand user intentions.

The West Lake Model differs — it possesses higher emotional intelligence, featuring long-term memory, personalization, and emotional perception. It can remember users' historical interaction information, deeply customize the model's persona, personality, and preferences, and perceive users' emotional states, interacting through warmer expressions.

Beyond high EQ, the West Lake Model also boasts high IQ.

According to internal evaluations, the West Lake Model has approached GPT-3.5 in multiple core tasks including writing and dialogue.

Xinchen AI internal research materials

The West Lake Model currently comprises over a dozen members, with parameter counts ranging from over 10 billion to over 100 billion.

Evaluation results from SuperCLUE in May and C-Eval in June this year show that the West Lake Model's IQ (cognitive and reasoning abilities) both rank among the top domestically.

SuperCLUE Evaluation Results (May 30, 2023)

C-Eval Evaluation Results (June 18, 2023)

03

Empowering Industries, Exploring the Unknown

At the West Lake Model launch event, Zhenzhong Lan stated: As models continue to grow larger, capabilities gradually strengthen, and training and inference costs significantly decrease, the iteration speed of large models will keep accelerating, scenario deployment will arrive faster than imagined, and they will be able to empower all industries.

The general-purpose large model with both high IQ and EQ built by Xinchen AI has considerable application potential across many industries, including entertainment (such as digital humans), gaming (such as intelligent NPCs), social (such as virtual companions), education (such as virtual teachers), healthcare (such as virtual doctors), and finance (such as intelligent investment advisors).

In these scenarios, what people need is not cold, tool-like AI, but intelligent agents with high-quality dialogue capabilities that are more human-like. These agents need to be deeply customizable and multimodal.

In terms of multimodality, Xinchen AI continues to achieve technical breakthroughs, such as more precise image and video generation.

Previously, Xinchen AI and Zhiyi Technology, an AI company deeply rooted in the apparel industry, jointly launched an AI large model for the fashion design industry — Fashion Diffusion. Designers only need to briefly describe clothing characteristics, and the model can generate high-quality realistic clothing effect images within 10 seconds.

FD Fashion Model intelligently generated styles

Artificial general intelligence is flourishing globally, gradually moving from perceiving the world to creating it.

Going forward, an increasing number of industries will require empowerment from general-purpose large models or vertical models, with application scenarios becoming ever richer.

The West Lake Model, with both high IQ and EQ, will continue to iterate, constantly expanding its capability boundaries, and becoming a world-leading artificial general intelligence platform.

---

👇For financing or cooperation opportunities, please long-press to scan the QR code below👇

Add BlueRun Ventures' contact on WeChat to connect with us directly.

▶ More Exciting Content ◀

Originating in Silicon Valley, BlueRun Ventures was established in 2005 as a venture capital firm focused on early-stage startups.

Currently, BlueRun Ventures manages multiple USD and RMB dual-currency funds in China, with assets under management exceeding 15 billion yuan, making it one of the largest early-stage funds domestically. Its investment stage focuses on Pre-A and Series A rounds, covering sectors including hard tech and innovative interaction, enterprise technology, new consumption, and healthcare. It has cumulatively invested in over 150 startup companies, including Li Auto, Waterdrop, QingCloud, Guazi.com, Qudian, Songguo Mobility, Ganji.com, Energy Monster, Yuntu Semiconductor, Machenike, Yunheng Intelligence, Anxin Network Shield, and BioMap.

BlueRun Ventures has been ranked first in Zero2IPO's "Top 30 Early-Stage Investment Institutions in China" and ChinaVenture's "Top 30 Best Early-Stage Venture Capital Institutions in China," and was named among Preqin's Top 10 consistently high-performing VC fund managers globally.

Additionally, BlueRun Ventures has consecutively received honors from Forbes China, 36Kr, Cyzone, Caixin Media, CBNweekly, Jiemian, and other media institutions, including "China's Best Early-Stage Institution of the Year," "China's Top Venture Capital Institution," "Most Entrepreneur-Friendly Early-Stage Institution of the Year," and "Most Influential Early-Stage Institution of the Year."