SynDexy recently announced that SynDexy C1, its in-house dexterous-manipulation humanoid robot, has completed its first small-batch delivery. The robots have been installed in factories run by customers in consumer electronics and electronic components, where they handle precision assembly work.

The key detail here isn't the delivery, it's the precision assembly. That step demands extremely high accuracy and stability, and it has been one of the hardest places for humanoid robots to break into over the past few years.

Getting a Feel for It Comes Before Assembly

SynDexy C1's headline capability is visuotactile sensing.

On the hardware side, it carries an in-house visuotactile sensor with micron-level spatial resolution and high-dynamic fine-grained sensing. Put plainly, the sensor has to tell how much force is being applied across a small patch of fingertip and where that force is distributed, not merely register that it has bumped into something.

On the model side, it runs a manipulation model called SynDexy M1, which the company describes as the world's first fusion of three states, vision, touch and force, combining all three streams of information.

Why add that layer? Vision tells you where something is and what shape it is. But assembly usually breaks down over questions vision can't answer: how much force went into the press-fit, whether the part slipped, whether the clip actually seated. A camera can't see any of that; only contact sensing can. Vision plus touch plus force lets the robot judge what is happening in its hand even when the contact surface is hidden.

At the system level, SynDexy C1 has 22 degrees of freedom across its body and runs motion control at 1-millisecond real-time rates. One millisecond means the control loop closes a thousand times a second. When a part drifts or jams mid-assembly, the system has to react on that timescale, or an insertion turns into a collision.

Over 98% Success Rate on Key Tasks in Earlier Testing

According to the company, SynDexy C1 exceeded a 98% success rate on key tasks in earlier testing.

A 98% figure is unremarkable in a grasping demo video. It means something different on an assembly step that has to run continuously on a production line. A line doesn't care about your best single attempt; it cares how many of several hundred attempts fail.

Six Months From Project Launch to Delivery

The pace is what makes this delivery notable: development began in April 2026, the first-generation prototype was finished in July, and small-batch delivery landed in October. Six months, start to finish.

In the humanoid robot industry, getting from prototype to factory floor normally means clearing several gates: hardware reliability, sensor calibration, process adaptation and customer acceptance. Covering all of it in six months says two things. First, the product was defined on the production line rather than built as a general-purpose platform in search of an application. Second, this generation has a deliberately narrow target: high-precision dexterous manipulation only, with no attempt to do a bit of everything.

The company sums up the approach as coming from the scenario and going into the production line, with real production-line requirements treated as hard criteria from product definition through technical development.

A Company Founded in 2025

SynDexy was founded in 2025 and is headquartered in Beijing. It positions itself as a full-stack embodied AI company focused on high-precision dexterous industrial manipulation, targeting consumer electronics, semiconductor micro-assembly, auto components and other high-precision manufacturing settings.

Full stack, in this line of work, means building three things in-house: the robot body, the visuotactile sensing and the embodied manipulation model. Those three are precisely where dexterous manipulation tends to come apart: the hardware moves well but the sensing falls short, or the sensing is fine-grained but the model can't use it.

On the team: chief scientist Fang Bin is a professor at the School of Artificial Intelligence at Beijing University of Posts and Telecommunications. The core R&D team comes from Tsinghua University, Beihang University, Harbin Institute of Technology and BUPT, along with tech companies including Huawei, DJI and Tencent. R&D staff account for more than 90% of headcount.

This delivery shows SynDexy has picked a very specific entry point: rather than building out general capabilities, it is pushing one thing, dexterous manipulation, to the point where it can go onto a production line. It's a narrow path, but every step gets tested by line acceptance.