YuanCe Future, an embodied intelligence algorithm startup, has unveiled Anchor 1, its whole-body control model. Positioned as the S0-layer "cerebellum" within the humanoid robot whole-body intelligence architecture, Anchor 1 is designed to deliver precise and stable whole-body motion output. The model rests on three core design principles: sparse commands anchored to five key points—the pelvis, left and right wrists, and left and right ankles; training data drawn from a rigorously curated dataset of million-level paired motions; and independent error feedback decoupled from motion encoding, which substantially enhances precision and stability. The team has also developed a dedicated evaluation framework spanning three scenario types: whole-body coordination, dynamic motion tracking, and pose reachability. Real-robot tests demonstrate that global position errors at the wrists and ankles are approximately 28% and 39% lower, respectively, than those of the best-performing comparison method. The efficiency of teleoperation demonstration collection is greatly improved, and the model can effectively support autonomous task execution by upper-layer large models. Currently, the industry's R&D focus is shifting from the humanoid robot "brain" to the "cerebellum," and from performance orientation to deployment. YuanCe Future plans to release Anchor 1's relevant weights, code, teleoperation collection pipeline, and evaluation tools in stages, promoting industry-wide adoption of precise whole-body manipulation capabilities for humanoid robots.