Focusing on interpreting global artificial intelligence technology and embodied applications | Bringing you a glimpse of the futureFollow me,here, you will see the ‘next step’ that ordinary tech media cannot show.Major benefit! The International Federation of Robotics (IFR) is offering the original PDF of the “2025 World Robotics Report” for free. Gain insights into global robotics trends; this report is all you need. Follow our public account to directly access authoritative materials.
Recently, the Swiss #robotics field has witnessed a significant breakthrough. Zurich-based startup Flexion Robotics AG has announced the completion of a $50 million Series A funding round. Founded in 2024, the company is dedicated to reconstructing the foundational paradigm of #robot development using generative artificial intelligence.

Unlike traditional scripting programming methods, Flexion has built a relatively complete three-layer technical architecture. At the command layer, large language models automatically decompose tasks described in natural language into executable subtasks, endowing #robots with common-sense reasoning capabilities; the motion layer processes specific action execution through a vision-language-action model primarily trained on synthetic data; while the control layer provides low-latency full-body control and a modular skill library to facilitate the rapid combination of new behaviors.
The launch of the Flexion Reflect v0 system also marks a critical step towards general autonomy. This system is based on agents of language models and vision-language models, capable of understanding daily routines and decomposing task objectives. Users only need to set the desired outcome through prompts and fine-tuning, and the system can automatically generate local trajectories with collision awareness, guiding the #robot to complete various actions from grasping objects to full-body navigation.

What is particularly noteworthy is the unique technical path. Flexion adopts a modular design to avoid the limitations of an “end-to-end monolithic architecture,” enhancing the overall system’s versatility by maintaining interface simplicity. In terms of data strategy, the company prioritizes the use of simulated and synthetic data, only selectively integrating real data to address specific deficiencies. This asymmetric data usage method significantly reduces training costs.
This funding round features a luxurious lineup, including top investment institutions such as DST Global Partners, NVentures (NVIDIA’s venture capital arm), and redalpine. Notably, this comes just months after the company completed a $7.35 million seed funding round, demonstrating strong recognition from the capital market for its technological approach.
The funding will primarily be used to expand the R&D team in Zurich, enhance computing and #robot clusters, and establish a branch in the United States to accelerate the commercialization of the technology stack. The company has already established partnerships with several large OEM manufacturers, and this new round of funding will support the global expansion of these collaborations.

🔎
#Generative AI is reshaping the trajectory of #robotics technology, shifting from reliance on precise programming to understanding-based autonomous decision-making. Flexion’s three-layer architecture demonstrates a viable development path for general-purpose robots, with its modular design and data strategy providing important references for the industry. As the capabilities of large models continue to evolve, robots are expected to break through the limitations of specific tasks and move towards true general intelligence. This transition requires not only algorithmic innovation but also a rethinking of how robots interact with their environment.
When robots can understand natural language and autonomously decompose tasks, which industries do you think will be most directly impacted? How should we construct a matching safety and ethical framework? We look forward to your forward-thinking insights.
➡️ Statement and Interaction / This article is translated from publicly available technical materials overseas, for discussion purposes only.
➡️ Extended Experience / For dynamic demonstrations, please follow the related video account 【Xiwangzhe】
If this article has inspired you, please like 👍 and ❤️ to support us!
Please open in the WeChat client