模思智能 Raises Over 1 Billion Yuan in Pre-A for a Real-Time Omni-Modal Model

Shanghai startup 模思智能 announced on October 12 that it has closed a Pre-A round of over 1 billion yuan. At the current exchange rate, that is about $140 million. 华控基金 led the round. The company says it will put the money into a foundation model for real-time omni-modal (全模态) interaction. Such a system is supposed to perceive and produce text, voice, images and video in one continuous stream, with no pause between "heard" and "replied."

The amount stands out even in a market where large early-stage rounds have long been common. A Pre-A round is usually measured in tens of millions of dollars, and this one is an order of magnitude larger. Investors are effectively funding the development of a proprietary foundation model, not an application product built on top of someone else's.

Who invested

The list of participants says as much about the round as the figure itself. Besides lead investor 华控基金, the deal includes 国开科创, 上海科创基金, 新微资本, 徐汇资本, 追光硬科技创投 and 临港策源基金. Most of them are tied to state or regional capital, primarily Shanghai's. Existing shareholders 元禾控股, 上海国投 (via 上海科创集团) and 星连资本 invested again. 多维资本 has been appointed exclusive financial advisor for the round and for the period ahead.

Repeat participation by earlier investors is usually read as a sign that the team delivered on the promises of the previous stage. But the company has not publicly disclosed what those results were.

What is known about the purpose of the round

The stated direction is a "base" (基座) for real-time omni-modal interaction. In China, the term refers to a foundation model on which other companies build applications. The key phrase here is "real time." The classic pipeline of speech recognition, then a language model, then speech synthesis accumulates latency at every step. A single model that works with several modalities directly can respond faster and stop mid-sentence when the user interrupts. Large labs are now moving in the same direction.

The original announcement, in the form available, breaks off at the description of goals. It says nothing about architecture, model size, benchmarks or release timing. Anything beyond what was stated therefore remains speculation.

What is confirmed:

• Round: Pre-A, over 1 billion yuan (about $140 million).
• Lead investor: 华控基金.
• New participants: 国开科创, 上海科创基金, 新微资本, 徐汇资本, 追光硬科技创投, 临港策源基金.
• Existing shareholders: 元禾控股, 上海科创集团, 星连资本.
• Financial advisor: 多维资本 (exclusive, for the round and for the long term).
• Use of funds: a real-time omni-modal interaction model.

What to watch next

For those following the Chinese market, the nearest signs of the project's maturity are fairly concrete. First: whether a technical report or open weights appear, which would allow response latency and speech quality to be assessed. Second: whether the company names its first customers, for example in voice assistants, smart devices or enterprise services. Third: how the round affects hiring, since real-time models require both researchers and expensive compute.

Developers and product teams should simply add 模思智能 to a watch list for now. Without an API, documentation and hard numbers, it is too early to build it into an architecture. Investors and analysts may find it useful to compare the terms of the round with those of other Chinese players in the multimodal segment. If the next steps confirm the ambitions, the Shanghai cluster will gain another contender for the role of base layer for voice and video interfaces.