French robotics-focused startup Genesis AI has released its first foundational model, GENE-26.5, alongside a human-like hand built in-house. The company says the hand is capable of carrying out complex tasks such as cooking, solving a Rubik’s Cube, preparing a smoothie, and playing the piano. The release marks an important step for the startup, which began with the goal of powering real-world robots with advanced AI rather than simply building another software-only model.

Key facts at a glance

• Genesis AI released its first foundational model, GENE-26.5.
• The company also built a human-like robotic hand in-house.
• The hand can perform tasks such as cooking, solving a Rubik’s Cube, preparing a smoothie, and playing the piano.
• Genesis AI completed a $105 million seed funding round in July last year.
• Co-founder and CEO Zhou Xian leads the company.
• Co-founder and president Theophile Gervet previously worked at Mistral.
• The company has offices in San Carlos, California; Paris; and London, with a team of about 60 people.
• Genesis AI is working toward a full-body, general-purpose robot.

The startup completed a $105 million seed funding round in July last year. That round was co-led by Eclipse and Khosla Ventures, with additional backing from Bpifrance, HSG, former Google chief Eric Schmidt, telecoms entrepreneur Xavier Niel, MIT Computer Science and Artificial Intelligence Laboratory director Daniela Rus, and Apple distinguished scientist Vladlen Koltun. With that backing, Genesis AI has moved quickly from concept to hardware.

From software-first strategy to in-house hardware

Genesis AI began with a focus on a model designed to power real-world robotics. Co-founder and chief executive Zhou Xian explained that the company then decided to build its own hand in order to have more accurate control over the full stack. By designing both the software and the hardware, Genesis AI can optimize how the model perceives, plans, and acts in physical environments. That vertical integration is a notable choice at a time when many AI startups prefer to license models and avoid the complexities of manufacturing.

The hand itself is designed to closely resemble a human appendage. This is a deliberate decision. Many robotic hands built by other companies have two or three fingers, which simplifies gripping but limits dexterity. Genesis AI’s human-like form allows data to be more directly transferred to the device, meaning demonstrations and training data captured from human hands can be mapped onto the robot with less distortion. The company argues that this anatomical similarity is key to making the robot useful in environments built for people.

GENE-26.5: a foundation model for physical tasks

The model, GENE-26.5, is the first foundational model from the company. Foundation models in robotics are typically trained on large and diverse datasets, then adapted for specific tasks. In this case, GENE-26.5 is designed to translate visual and sensory inputs into precise physical actions. The name appears to reference both the generation of the model and the number of parameters, although the company has not publicly confirmed the parameter count.

Robotic foundation models have become a crowded field. Companies such as Google DeepMind, Figure AI, and others are experimenting with large language models and vision-language-action models to help robots understand commands and manipulate objects. Genesis AI is entering that space with a hardware-first twist: the same team can alter the physical hand to improve the model, and the model can inform the design of the next-generation hand.

One of the biggest challenges in robotics is generalization. A robot trained to pick up one object may fail with another object of a different shape, in different lighting, or in a cluttered environment. Genesis AI hopes that by using human-like hands, it can collect a richer set of training data and make it easier for the model to generalize from human videos and sensor recordings. The company has also created lightweight sensor gloves specifically for this purpose.

Delicate tasks and data collection

Genesis has developed lightweight sensor gloves that can be worn by workers in fields such as pharmaceuticals or manufacturing. These gloves capture detailed hand movements and pressures, which can then be passed along to the AI model. The company says data from the gloves can be combined with information gleaned from videos of workers carrying out tasks. This multi-modal approach could accelerate training by capturing not just what the hand does, but also how the hand feels.

The choice of sensor gloves raises an important question about workplace acceptance. Workers may be reluctant to help train a robotic replacement that could someday take over parts of their jobs. Genesis AI has not disclosed how it plans to address these concerns, but the use of existing workers’ expertise is a common strategy in robotics. Many companies argue that human workers will remain essential for years and that robots will handle repetitive, dangerous, or physically demanding aspects of a job.

In a demonstration video, Genesis AI showed a pair of its hands carrying out complex or delicate processes. The hands prepared a smoothie, played the piano, solved a Rubik’s Cube, and performed other tasks that require fine motor control. These demonstrations are intended to showcase the precision and adaptability of the system, although it remains to be seen how reliably the hands perform outside controlled settings.

Team and global footprint

Co-founder and president Theophile Gervet previously worked for France’s Mistral, a well-known European AI company. That background gives Genesis AI a strong connection to the European AI ecosystem. The company says it retains a substantial presence in Europe because of the continent’s large talent base and its potential market of industrial customers.

Genesis AI’s team of 60 people is split roughly evenly between the United States and Europe, with a slightly larger group at its US base in San Carlos, California. The company also has offices in Paris and London. This transatlantic structure allows it to tap into both American capital and deep-tech expertise and European engineering talent and industrial relationships.

The company is now working to supplement the mechanical hand with a full-body, general-purpose robot. The hand is likely the first component of a larger system that could eventually include arms, a torso, legs, or a mobile base. A full-body robot would be able to navigate around work environments and perform a wider range of tasks, rather than operating only as a hand mounted on a fixed rig.

Investor confidence and market context

The $105 million seed round was unusually large for a seed-stage company. Seed rounds of that size are more common in AI infrastructure or robotics, where capital-intensive hardware and long development timelines require substantial funding. The participation of prominent investors including Eric Schmidt, Xavier Niel, Daniela Rus, and Vladlen Koltun signals confidence in the team’s ability to deliver.

Vladlen Koltun, previously at Apple and known for his work in machine learning, and Daniela Rus, director of MIT CSAIL, bring considerable academic and industrial credibility. Their involvement may help Genesis AI attract researchers and form partnerships with universities and labs. Eric Schmidt, former CEO of Google, has invested in a wide range of AI ventures and often focuses on companies with transformative potential.

The robotics market has seen a surge of investment in recent years. Improvements in AI models, sensors, actuators, and battery technology have made humanoid and general-purpose robots more plausible. At the same time, many companies remain at the prototype stage, and widespread deployment in factories, warehouses, and homes is still limited. Genesis AI will need to prove that its human-like hand is not only impressive in demonstrations, but also reliable, durable, and cost-effective at scale.

European industrial opportunity

Europe is a particularly important target for industrial robotics. The continent has a strong manufacturing base in Germany, France, and Italy, as well as a growing need to address labor shortages and aging workforces. Genesis AI’s presence in Paris and London gives it access to these markets. The company believes that industrial customers will be early adopters of robots that can handle delicate tasks such as assembling electronics, sorting medical devices, or packaging pharmaceutical products.

However, the use of AI-powered robots in the workplace raises regulatory and ethical questions. The European Union has been drafting rules for AI and robotics, emphasizing safety, transparency, and human oversight. Genesis AI may need to align its development process with these regulations as it moves from research to commercial deployment. The company did not provide a timeline for shipping products or announce a price for the hand or the full-body robot.

Challenges ahead

Genesis AI’s success will depend on several factors. The model must perform robustly in real-world settings, where lighting, objects, and people vary. The hand must be mechanically durable, easy to maintain, and safe to operate near humans. The company must also demonstrate that its sensor gloves and data pipelines can provide a sustainable advantage over competitors.

Another challenge is computing cost. Foundation models for robotics often require massive datasets and significant compute to train. Genesis AI has not disclosed how it trains GENE-26.5 or what infrastructure it uses, but the $105 million seed round may be used partly for GPU clusters and data collection. The company will also need to hire roboticists, machine learning engineers, and manufacturing experts as it expands.

Competition in the space is intense. OpenAI has invested in robotics companies, Figure AI has raised billions, and established manufacturers such as Boston Dynamics continue to push the boundaries of humanoid robots. Genesis AI is differentiated by its focus on a human-like hand and its willingness to build hardware in-house, but it will need to execute quickly to avoid being overtaken by larger rivals.

The company’s approach of combining a foundation model with a human-like hand could help it stand out in a crowded market. If the hand can perform complex tasks with enough reliability, it could become the basis for a broad platform for future robots. The company says it is working to supplement the mechanical hand with a full-body, general-purpose robot, and the team’s experience in both Europe and the US gives it a solid foundation for that effort.

Source: Silicon UK News