Daimon Robotics has launched Daimon-TWM, which Robotics & Automation News describes as a tactile-grounded world model. The company says the model is meant to help robots interpret force, friction, deformation, and slippage, placing contact-related conditions alongside the information used during manipulation. Daimon’s official English site identifies the company as Daimon (Shenzhen) Robotics Technology Co., Ltd. and presents Daimon-TWM as part of its approach to physical AI. The launch is a company development with a stated technical purpose; the available account does not provide independent performance results.
That stated purpose is tied to a stack Daimon describes as visual tactile sensing, real-world physical-interaction datasets, and a tactile physical-world foundation model. The company says its data infrastructure is designed to keep expanding real-world interaction data for model iteration. Separately, 36Kr Europe reported on August 11 that Daimon had completed a strategic financing round of several hundred million yuan led by Ant Group, following a 100-million-yuan Series A that it said had closed two months earlier. The report did not specify the strategic round’s exact amount, valuation, ownership stakes, or allocation among investors.
Daimon-TWM centers contact information in robot learning
Daimon-TWM puts a particular emphasis on what happens when a robot makes physical contact. In Daimon’s description, a robot needs to account for force, friction, deformation, and slippage rather than draw chiefly on visual information. Those conditions frame the launch as an effort to make contact itself more central to the information a robot uses while handling objects. Daimon says the model connects an understanding of physical conditions with predictive choices and immediate control, in service of precise, adaptable, and dexterous manipulation in complex environments.
The distinction matters within the company’s own technical framing because the inputs it identifies extend beyond a scene’s visible appearance. The company’s stated three-part stack joins sensing, data collected from physical interactions, and a tactile foundation model. Daimon also says its dataset infrastructure continuously expands interaction data to support model iteration. These are descriptions from Daimon of how it is organizing its technology and data program, rather than disclosed results from an outside test.
EastFrontier has separately reported on Mifeng’s embodied-AI data platform funding and 51World’s embodied-AI training data platform. Those historical articles are background context and do not support Daimon-TWM’s technical claims.
Robotics & Automation News says Daimon officially began operations in 2023, is headquartered in Shenzhen, and maintains an R&D operation in Hong Kong. The outlet also reports that the company originated from a robotics research team at the Hong Kong University of Science and Technology. That background places the model launch within a company focused on a robotics problem defined by physical contact, with its formal English company name listed on its own site as Daimon (Shenzhen) Robotics Technology Co., Ltd.
VTLA combines visual and tactile information for manipulation
The new model builds on Daimon’s Vision-Tactile-Language-Action, or VTLA, architecture, according to Robotics & Automation News. The outlet associates Daimon-TWM with demonstrations involving glass, fruit skewers, and object organization in a refrigerator. They are demonstrations associated with the model, not disclosed deployments or customer results. The account also does not provide a test protocol, task-success figures, or an independent evaluation for them.
The company’s stated visual-tactile orientation links its model, sensors, and data collection. The reported glass, fruit-skewer, and refrigerator scenes are demonstrations shown with Daimon-TWM, while the available account does not provide independent measurement results for them.
Daimon says its physical-interaction datasets are an input to ongoing model iteration, alongside tactile sensing and its tactile physical-world foundation model. The available materials describe the company’s own data program in those terms.
Reported funding provides context for Daimon’s data program
36Kr’s August financing account provides a separate view of Daimon’s current company context. It reported that Ant Group led the strategic round of several hundred million yuan and that existing shareholders followed on. Among the industrial-capital backers cited by 36Kr are China Mobile, China Telecom, Inovance Industrial Investment, China Merchants Capital, and Lenovo Capital and Incubator Group. Its account does not establish that every listed party participated in the August strategic round, so the lineup should not be read as a roster of new co-leads.
The earlier Series A and the later strategic financing are distinct events in 36Kr’s account. The 100-million-yuan figure belongs to the Series A identified as having closed two months before the August report. Ant Group’s lead role applies to the subsequently reported strategic round. Keeping those details separate is important because the report supplies neither a precise amount for the strategic financing nor a valuation or investor-by-investor allocation.
Daimon says it is expanding real-world interaction data for model iteration, and the financing report describes strategic capital around the company. The funding report does not establish a specific effect on Daimon-TWM’s technical performance.
Daimon-TWM has been presented as a tactile-grounded model built on VTLA, with demonstrations involving glass, fruit skewers, and refrigerator organization. The available material characterizes those examples as demonstrations and does not provide an independent task-performance evaluation.