---
title: "Westlake Xinchen Raises Strategic Round from Tom Cat & BlueRun for Multimodal LMs"
author: "Unique Research"
sourcePublication: "Unique Research (FFCAP English library)"
originalPublishedAt: "2023-04-28T15:57:21+08:00"
canonical: "https://ffcap.cn/en/research/20230428a05wcu00"
source: "https://ffcap.cn/en/research/20230428a05wcu00"
language: "en"
cnSourceUrl: "https://view.inews.qq.com/a/20230428A05WCU00"
---

# Westlake Xinchen Raises Strategic Round from Tom Cat & BlueRun for Multimodal LMs

_Unique Research · Company / funding · April 28, 2023_

On **Apr 21, 2023**, AI company **Westlake Xinchen** (西湖心辰) closed a strategic round led by **Tom Cat** (汤姆猫) with **BlueRun Ventures** (蓝驰创投) participating — second raise in ~a month after **Mar 27** multi-ten-million USD Pre-A from Baidu Ventures, Dongkaitai, and Westlake University Education Foundation sustainability platform. This round: multimodal LM R&D + business expansion; Tom Cat becomes largest external shareholder outside the team. Prior round: team build + multimodal LM acceleration.

Founded **2021** by **Lan Zhenzhong**’s team with Westlake University AI research partners — AIGC for content + psychological counseling commercialization. Products: AI paint, AI writing, AI counseling. Lan: CMU AI PhD; ex-Google AI scientist; first author of language model **ALBERT**; MIT TR35 Asia-Pacific 2021; joined Westlake University 2020 as deep-learning lab lead / PhD advisor. Multimodal LMs = path to understanding/reasoning/self-learning AGI — analogous to humans learning via text/image/video.

**Architecture:** model layer → mid layer → app layer. **RIO** language LM series + dual-modal image-text models (dozens of members, ~10B–100B+ params; bilingual CN/EN; tool use: search, KG, external data). Mid layer: domain models (e.g. marketing) — API or custom; unstructured enterprise data newly valuable. Apps: counseling AI “Xiaotian”; writing “Friday” (>400k users, partial monetization; dialog bot with web + multimodal I/O); image “Dream Diary” (>1M users) — early data flywheel. RIO tested ≈ GPT-3.5 class; 2023 plans two iterations targeting GPT-4 parity on some tasks.

Lan’s CMU multimedia/CV background (IARPA multimedia event detection champ vs Stanford et al.); Google work with Transformer authors on Talking-Heads Attention / ALBERT on 1024 TPUs (Google News/Assistant). New raise open; seeks talent + strategic/resource-aligned investors for AGI for all.

_4 CN images; attach on ingest. Author: Huajia; review Ski._

— END —

---

Editorial English edition published on UniqueCapital: https://ffcap.cn/en/research/20230428a05wcu00
On-site reading page: https://ffcap.cn/en/research/20230428a05wcu00