Physix Frontier · News Briefing Card (Leiphone · Nov 16, 2026)

Ant Lingbo Embodied Model Challenge Ends, TARS Team Wins First Prize with 82.5

KEY FACTS

  • On November 16, Ant Lingbo announced the final results of its first Embodied Large Model Challenge, with team TARS taking first prize.
  • Fifteen finalist teams ran a 48-hour real-robot hackathon in Shanghai, completing three tasks: grasping fruits and vegetables, grasping cosmetics, and dual-arm sweeping.
  • All teams used the same open-source model LingBot-VLA 2.0, identical hardware and compute, and the final featured three rounds of real-robot evaluation with the highest score counting.
  • Team TARS has an average age of 25, and its captain is Wu Haonan, a post-00s master's student at Tianjin University who is about to graduate.
  • The event drew 1,855 teams and 3,252 contestants from 13 countries, covering 648 universities and companies.
  • Ant Lingbo also announced a developer ecosystem plan the same day, aiming to invite 20 most valuable experts and recruit 40 developer experts.

KEY DATA

82.5 pointsTARS team final score
1,855 teamsNumber of registered teams
3,252 contestantsNumber of registered contestants
648 institutionsNumber of universities and institutions covered

PHYSIX OBSERVATION

With a unified model, unified hardware and unified tasks, this competition is more like a public controlled-variable experiment: failure boundaries and tuning experience are being deposited as reusable engineering assets. For the industry, real-robot data collection and transparent-object grasping remain hard nuts to crack; for developers, the ecosystem plan means the entry channel is opening, but the Scaling Law moment for embodied intelligence has not yet arrived.

Source: Leiphone report