AGIBOT unveiled GO-3 Preview on November 3, an embodied foundation model that secured first-place rankings on three major embodied AI benchmarks: RoboDojo, RoboTwin, and RoboColiseum2. The company said the model will be fully open-sourced by the end of 2026.
GO-3 Preview is built on AGIBOT's proprietary embodied-native architecture, WRAM (World Reasoning Action Model), which unifies task understanding, future-state prediction, and action generation to enable robots to generalize across tasks and perform precise manipulation. The model supports large-scale learning across what AGIBOT calls the full lifecycle of pre-training, post-training, and post-deployment.
Two additional systems complete the pipeline. Genie Evolver, AGIBOT's hybrid post-training system, and Genie Master, its embodied agent framework, together with GO-3 Preview form what AGIBOT describes as an integrated learning system spanning pre-training through real-world operation.
The model release follows a period of rapid scaling for AGIBOT. The company's general-purpose AI robots reached cumulative production of 20,000 units in the first half of 2026, generating revenue of nearly 2 billion yuan (approximately $300.7 million)[2]. AGIBOT has claimed a 43% share of the global humanoid robot market[2].
ANALYSIS Open-sourcing GO-3 Preview by year-end would place a benchmark-leading embodied foundation model into the public domain, potentially accelerating third-party development on AGIBOT's architecture and expanding the installed base for its hardware platform.