若愚科技

Visko is an AI research company building "Live Models" — world models that run continuously in real time rather than rendering static clips. Founded in 2025 in Sunnyvale, California, its 16-person team draws from Apple, Google DeepMind, Meta, Amazon and Tesla, advised by Michael I. Jordan (UC Berkeley), Steve WaiChing Sun (Columbia) and Mengye Ren (NYU).

Pre-Seed · $10M · 2026-09

Nomic, founded in 2022 in New York, released GPT4All, an open-source AI model available for free download and local execution on consumer devices. It also offers Atlas, a tool that helps users visualize and explore the unstructured datasets used to train large language models.

Undisclosed · $17M · 2023-07 · Valuation $100M

面向人机自然交流的多模态语音与音频大模型及智能体平台

A robotics brain company focused on embodied intelligence, developing a proprietary universal world model that enables a single system to adapt across humanoid, quadruped, wheeled, AGV, and autonomous delivery vehicle platforms. The technology gives robots the ability to move, navigate, and interact autonomously in real-world human environments. The core team largely comes from Baidu Apollo and Inceptio Technology.

Angel · $7M · 2026-06

Founded in January 2026 and headquartered in Beijing, Delta Intelligence develops native general-purpose Humanoid Foundation Models (HFMs) for whole-body loco-manipulation, building its own data-collection hardware and data closed-loop to bring embodied AI into real industrial and household scenarios.

Angel · 2026-07

Modal Transition (模态跃迁), founded in 2023 in Wuhan, China, is a developer of multimodal AI large models and AI Agent foundation infrastructure.

Angel · $5M · 2025-12

OctaiPipe was founded in 2016 in London, UK, and operates an end-to-end Federated Learning Operations (FL-Ops) platform for edge AI deployment, designed for data scientists and AI engineers in critical infrastructure sectors. The platform orchestrates distributed machine learning across IoT device networks and is available on Microsoft Azure, AWS, or private cloud, and has received a £500,000 grant from Innovate UK.

Undisclosed · $4M · 2024-01

Emulate is a London AI startup founded in August 2026 by three former Google DeepMind researchers behind the Genie world models, building AI world models that simulate how the physical world behaves for robotics simulation and interactive 3D environments. It remains in stealth with no public product or website.

Mistral AI, founded in 2023 in Paris, France, develops open-source LLMs (large language models) to compete with OpenAI.

Series D+ · 2026-09 · Valuation $24B

Gradient Network, founded in 2024 and headquartered in Singapore, is building a decentralized network that enables developers to run AI models without relying on cloud services.

Seed · $10M · 2025-06

Cohere, founded in 2019 in Toronto, Canada, provides enterprises with generative AI products including text models, embedding models, and retrieval models. The company serves hundreds of customers including Notion, Oracle, and Bamboo HR.

Undisclosed · $1.8B · 2025-09 · Valuation $5.5B

AI21 Labs, founded in 2017 in Tel Aviv, Israel, provides developers with customized text-based business applications powered by AI21 proprietary AI models. Main products include Wordtune (a Chrome extension for improving writing) and the LLM API AI21 Studio.

Series D+ · $508M · 2025-05 · Valuation $1.4B