Ant Lingbo draws 1855 teams to its first embodied AI challenge

The online qualifying round of the first Ant Lingbo Embodied Large Model Challenge ended on 27 October. The organizer, Ant Lingbo Technology, calls the result an industry record: 1855 teams and 3252 participants from 13 countries entered. For a first-ever tournament on embodied intelligence, that is a very large turnout. The competition now moves on to the finals.

Who showed up for the debut tournament

The organizers' official statistics:

• 1855 teams submitted entries;
• 3252 people registered as participants;
• 13 countries are represented;
• 648 organizations: well-known universities, research institutes, large technology companies and companies working on embodied intelligence;
• 41 foreign universities and organizations are among the participants.

Simple arithmetic gives an average team size of 1.75 people. Many teams signed up as pairs, and some, apparently, as individuals. That is typical for open AI competitions: the barrier to entry is low, so enthusiasts and small startup teams compete alongside lab groups.

Why the makeup of the field matters more than the headline number

The figure of 1855 teams grabs attention, but the structure is more interesting. The organizers put academic labs, corporate R&D units and dedicated embodied AI companies on a single list. These groups are usually compared indirectly, through papers, demo videos and press releases. On a shared platform with a single set of rules, their solutions can be placed side by side for the first time.

The presence of 41 foreign organizations also matters. Embodied AI, meaning large models that control robots in a physical environment, remains a field where Chinese and international teams work in parallel and shared benchmarks are scarce. A tournament of this size could become one, if the organizers prove that their evaluation is transparent and the results are reproducible.

What comes next and what to watch

The online stage is over, and the finals are ahead. That is where it will become clear how strong the finalist teams are, which approaches to embodied models prove more effective, and how many participants make it to the end.

For Ant Lingbo, the competition serves several purposes at once. It is a way to find talent, draw the community's attention to its own platforms, and get an independent assessment of what can be built on top of current models. Ant Group has long invested in AI, and the embodied direction is one of the most expensive and riskiest bets in the industry. A venue where thousands of engineers compete every year lowers the cost of finding ideas.

Teams and researchers working with robots and embodied models should follow the finals with three questions in mind: which approaches came out ahead, how the organizers assess quality on physical tasks, and whether the results will be made public. The answers will determine whether the tournament becomes an industry reference point or remains a large but one-off event.