Physix Frontier · News Briefing Card (TMTPost · Nov 18, 2026)
Ant Lingbo open-sources LingBot-Touch tactile model, launches 1.5B yuan first round
KEY FACTS
- On November 10, Ant Lingbo released and open-sourced LingBot-Touch, a tactile foundation model that unifies the representation of three types of tactile signals.
- The model was pretrained on nearly 5,000 hours of real tactile data, and its code, weights and technical report are all open-sourced.
- Across 26 real-robot tasks, plugging in the model raised the average success rate of four action models by 14 to 26 percentage points.
- Ant Lingbo launched its first external funding round in August 2026, aiming to raise 1.5 billion yuan; completion has not been disclosed.
- Its LingBot series has been open-source for nine months, drawing over 33,000 GitHub stars and more than 2.2 million downloads.
KEY DATA
nearly 5,000 hoursTactile pretraining data duration
14 to 26 percentage pointsAverage success rate improvement
1.5 billion yuanTarget raise for first external round
over 33,000GitHub stars
PHYSIX OBSERVATION
Touch is the threshold for robots doing fine-grained work, and an open-source foundation can pull in an ecosystem, but the data sits in the hands of dexterous-hand makers, and whether that supply can be sustained determines the model's ceiling. Ant Lingbo has the group behind it and no shortage of patience, but during an industry shakeout investors look harder at work-hour logs and repurchase rates; the 1.5 billion yuan first-round valuation will test the market's real confidence in the general-purpose brain route.
Source: TMTPost report
Physix Frontier