Octopus Power "Full Throttle"

In the embodied intelligence track, Octopus Dynamics is absolutely a one-of-a-kind player.

The team boasts a remarkable resume, capital pours in swiftly, and product advancement is impressively aggressive. Founded only in January this year and officially launched operations in March, it secured multiple rounds of financing from top-tier capitals within half a year. Horizon Robotics, Xiaomi, and GL Ventures all joined in, with total accumulated funding reaching nearly 1 billion yuan, and the latest 500 million yuan financing round is about to close.

"There is capital looking to invest in the company every month," an investor told 36Kr.

On the product front, Octopus Dynamics has placed simultaneous bets on the three hottest tracks in current embodied intelligence: world model, dexterous hand, and data collection. The company even hung a banner of "Decisive Battle at WRC" very early on, taking this conference as a clear R&D deadline.

The team lived up to expectations. Right during the 2026 World Robot Conference, Octopus Dynamics released three core products at once: SYNWorld V0.5, the embodied-native general world foundational model, OctoH-Hand, the high-degree-of-freedom bionic rope-driven dexterous hand, and OctoSense, the next-generation full-modal data acquisition system.

What is even more striking is that among a large number of embodied intelligence startups that survive on "external blood transfusion", Octopus Dynamics has taken the lead in achieving self-sustainability -- its commercial implementation performance is outstanding, and its revenue is expected to hit the 100 million yuan level by the end of 2026.

Who exactly is Octopus Dynamics? Or, in this increasingly "red ocean" embodied intelligence track, what makes it stand out from the crowd?

"Integrated Brain-Hand", Returning to the First Principle of Intelligence

Over the past two years, the embodied intelligence industry has become increasingly segmented. Some teams work on VLA, some on world models, some on dexterous hands, and more and more companies are venturing into the field of data collection.

For a startup, "focus" usually means that it is easier to achieve single-point breakthroughs and deliver results quickly.

However, for foundational models, dexterous hands, and data collection, Octopus Dynamics has almost simultaneously entered the three extremely competitive directions in the current embodied intelligence sector, taking a very "heavy" development path.

Why would a newly established company voluntarily extend its business front so widely?

Dou Dalong, Founder and CEO of Octopus Dynamics, gave his explanation based on his understanding of "intelligence" itself.

"Human intelligence has never evolved separately from the body. As hands become more dexterous, humans gain more complex operation capabilities; as tools become more sophisticated, they in turn drive the continuous development of the brain," said Dou Dalong.

The "smart brain" and "dexterous hand" are inherently complementary and mutually reinforcing.

At this point, the first keyword to understand Octopus Dynamics comes to the surface -- "Integrated Brain-Hand".

Let's start with the "brain" first.

The SYNWorld series foundational large model is at the most underlying layer. The newly released SYNWorld V0.5 is an embodied-native general world foundational model with over 10B parameters, trained on more than 300,000 hours of data. Its MoT architecture conducts cross-modal modeling through shared attention at the same time, bringing text, vision, motion, force sense, tactile sense and proprioceptive state into a unified framework -- which is also the core idea of Octopus Dynamics' "vision-force-touch" technical path.

Based on SYNWorld, the model expands in two directions: ACWM (Action-Conditioned World Model) and WAM (World Action Model).

ACWM focuses on "what will happen next if the robot executes this action"; WAM generates the actions that the robot should take according to the environment and goals. The former is more oriented to world prediction, while the latter is more oriented to action decision-making, and both share the understanding of the physical world.

Through this unified training, the two separate systems of "understanding the world" and "acting on the world" for embodied intelligence are no longer isolated, thus achieving true generality.

This technical path is not exclusive to Octopus Dynamics. At present, Google DeepMind, NVIDIA and many other industry giants are all advancing along similar directions.

However, the particularly outstanding breakthrough of Octopus Dynamics is the continuous scaling capability of the SYNWorld model.

Continuous scaling means that as data volume, model scale, computing power and task coverage keep growing, the model's capabilities can also evolve continuously.

At present, Octopus Dynamics has clearly positioned the current SYNWorld V0.5 as the starting point of this scaling curve. With parameters over 10B at the current stage, the company plans to continue expanding to over 20B, and then advance to hundreds of billions of MoE, while gradually extending the evaluation scope to more tasks, scenarios and robot bodies.

What is particularly noteworthy is that Octopus Dynamics has built a series of infrastructures covering data, training, inference and evaluation around the SYNWorld foundational model series. Among them, the Data Infra has a daily average data production capacity of over 10,000 hours, and the actual measured training efficiency of the Training Infra is more than 10 times higher than that of similar open-source achievements.

This exactly leads to the second product -- the OctoSense full-modal data acquisition solution.

In today's embodied industry, one of the universal consensuses is the severe shortage of data. Especially for the "tactile sense" segment in the "vision-force-touch" track that Octopus Dynamics focuses on, data is extremely scarce.

Fan Qingyuan, Co-founder of Octopus Dynamics, revealed to 36Kr that the OctoSense data acquisition solution was originally only for internal use by the Octopus Dynamics team. However, due to the lack of mature solutions in the industry, a large number of customers and partners have actively come to discuss cooperation after the product was launched in a commercial form.

Specifically, the OctoSense full-modal data acquisition solution of Octopus Dynamics includes three core products: a four-fisheye Ego headband, an EMG wristband, and a bionic exoskeleton data glove.

The three products fill three gaps respectively. The Ego headband addresses the needs of large-scale data collection and first-person perspective vision; the exoskeleton glove complements high-precision hand pose, contact and force information; the EMG product further solves the problems of visual occlusion and unobservable active force exertion.

Among them, the EMG wristband has drawn widespread attention from the industry.

Compared with vision or force sensor solutions, EMG signal collection is more direct and more real-time. It is not affected by visual occlusion, and is lightweight enough to greatly lower the threshold for collecting high-quality human demonstration data -- after all, the difficulty of an assembly line worker wearing a bulky data glove to work for a whole day is incomparable to that of wearing a lightweight EMG wristband to work for a whole day.

However, the prominent scaling obstacle of traditional EMG technology is that EMG signals naturally have inter-individual differences. Sensor position, sweating, and fatigue will all change the signal distribution, and re-calibration is required when switching to a different user.

The remarkable breakthrough of Octopus Dynamics is that it has built the SynEMG EMG large model, taking the lead in realizing zero-shot generalization of EMG signal collection across individuals. When facing new individuals not included in the training set, the model does not need to add new individual data or conduct incremental training, realizing the transformation from "one model for one person" to "one model for thousands of people".

Solving the generalization problem is equivalent to breaking through the ceiling that limits traditional EMG collection solutions, making the scaling of embodied intelligence data collection possible.

The third product, the OctoH-Hand dexterous hand, is the other end of Octopus Dynamics' "Integrated Brain-Hand" idea: enabling embodied intelligence to truly enter the physical world.

The OctoH-Hand adopts a high-degree-of-freedom rope-driven structure with 23 active degrees of freedom and 28 independently controllable actuators. More than 1900 tactile sensing units are integrated in the palm and fingers, and full closed-loop force control is constructed through joint current feedback and tactile feedback, with the Mean Time Between Failures (MTBF) reaching 1000 hours.

There are three key designs here:

First, and the most groundbreaking one, is the "digital twin" design concept of the OctoH-Hand.

The OctoH-Hand dexterous hand and the exoskeleton data glove in the OctoSense data acquisition solution adopt a one-to-one corresponding motion configuration, degree of freedom definition and perception configuration.

In traditional solutions, the data collected by humans usually needs to go through motion retargeting, parameter mapping and debugging before it can be converted into actions that the robotic hand can execute.

The solution of Octopus Dynamics enables the collected data to be directly imported into the dexterous hand as much as possible, realizing "collection for immediate use", opening up the loop between "brain-hand-data", and turning the dexterous hand from an independent hardware into a part of the entire physical AI system.

Second, it is highly bionic. The actuators of OctoH-Hand are placed at the back of the forearm, and then drive the fingers through tendons, which not only resembles the human muscle-tendon structure, but also accommodates more degrees of freedom in the limited palm space. Its dexterity, motion capability and operation capability are very close to those of human hands.

Third, it is full closed-loop force control. After robots truly enter household, industrial and service scenarios, the core difficulty of operation will soon upgrade from "grasping objects" to "operating with appropriate force". Based on the back-drivable electromechanical characteristics of joints, OctoH-Hand integrates joint current loop feedback and dual force perception signals of tactile sense, and builds a real full closed-loop force control system from internal driving force to external contact force, which can complete extremely fine dexterous hand operations such as screwing, inserting, pinching, rotating, and pulling.

The three products, namely foundational model, dexterous hand and data acquisition, form a complete closed loop: humans generate data, the model learns physical laws; the model generates actions, the dexterous hand executes in the real environment; new tactile, force sense and task results are generated during the execution process, which then flow back to the next round of training.

Octopus Dynamics calls this entire set of ideas Bio2Robot. It tries to convert human biological operation information such as vision, motion, contact, and force exertion into data that robots can learn and reuse, then the world model learns the laws from the data, and the dexterous hand returns to the real world.

Extreme Scaling, Extreme Generality

During the communication between 36Kr and the founding team of Octopus Dynamics, one word appeared repeatedly -- "carbon-based data".

How to convert the constantly occurring motion experience in real human labor into training data that machines can use, build smarter and more general robot brains, and promote the realization of "silicon-based intelligence".

Yes, the real ambition of Octopus Dynamics is not just to build a smarter robot and let it go to the factory to "do screw tightening work".

What Octopus Dynamics really wants to build is "general physical intelligence" that can achieve extreme scaling and extreme generality.

In other words, it is embodied intelligence AGI.

At this point, the second keyword to understand Octopus Dynamics comes to the surface -- "Extreme Scaling, Extreme Generality".

In the past six months, the embodied industry has been talking about Scaling Law more and more frequently. VLA has proved that robots can learn more and more tasks, but tokens in the language world are naturally unified, and the basic semantics of a word will not change drastically no matter where it appears; the physical world is highly continuous, and the same action of "grabbing a cup" will have different results due to changes in the material, weight, friction coefficient of the cup, the height of the tabletop, and the hand shape of the robot.

This is also why there is still no universally recognized embodied intelligence Scaling Law in the true sense in the industry so far.

On the path of scaling, Octopus Dynamics has a very radical idea.

The reason why it works on foundational models, dexterous hands and data acquisition at the same time is that it hopes data, models and hardware can all continue to expand, and form a positive flywheel among the three: larger volume of data trains stronger models, stronger models are applied in more real tasks, real tasks generate more high-quality data, which is then used to train the next generation of models.

Dou Dalong once said that if the possibility of scaling is confirmed, the company will "go all out to achieve scaling at all costs".

Accompanied by "extreme scaling" is the "extreme generality" of implementation scenarios.

At present, a large number of embodied intelligence capabilities still highly depend on specific tasks, specific scenarios and specific robot bodies. The "one scenario, one training" that is often mentioned in the industry is essentially an extension of "special machine for special use" in the traditional automation era.

However, what Octopus Dynamics pursues is not to make robots perform extremely well in a fixed workstation, but the huge generality potential brought by model scaling and data scaling.

Industry, commerce, household; manufacturing, scientific research, companionship... All the far-reaching technological revolutions in history are accompanied by the large-scale invocation of a certain scarce capability. The steam engine amplified physical strength, electricity reduced the cost of energy use, and computers greatly improved information processing capabilities.

Dou Dalong believes that if robots become an important infrastructure for the next round of productivity revolution, what they really need to release is physical operation capability.

And this is exactly the real ambition of Octopus Dynamics.

Team Members "With Light in Their Eyes"

Finally, if you want to truly understand Octopus Dynamics, the "team" is the most indispensable factor.

Dou Dalong, Founder and CEO, participated in early deep learning R&D at Baidu IDL, then became the 6th employee of Horizon Robotics, participating in BPU development and AI engineering; in 2021, he co-founded Zhiji Robotics and served as CTO, promoting autonomous driving technology to mass production for real car manufacturers. At the end of 2025, Zhiji was acquired by NavInfo.

The value of this experience has become even more prominent in the era of embodied intelligence.

At the global first press conference successfully held by Octopus Dynamics at WRC on August 19, there is a detail closely related to the "team".

Dou Dalong specially thanked his "two former bosses" -- Yu Kai, Founder and CEO of Horizon Robotics, and Cheng Peng, CEO of NavInfo, and their ways of supporting Octopus Dynamics are different.

Yu Kai's support is clearly reflected in his status as a shareholder, while Cheng Peng's support was implemented as a strategic cooperation agreement signed the day after the press conference, which more directly connects to Octopus Dynamics' product implementation and industrialization path.

On August 20, Octopus Dynamics and NavInfo officially announced the launch of