{"id":"agibot-go-3-preview","slug":"agibot-go-3-preview","name":"GO-3 Preview","orgId":"agibot","family":"agibot-go-3-preview","releaseDate":null,"availability":"announced","openWeights":false,"license":null,"paramsTotal":null,"paramsSource":null,"contextWindow":null,"knowledgeCutoff":null,"modelCardUrl":"https://www.agibot.com/article/231/detail/131.html","hfRepo":null,"modalities":["robotics"],"aliases":["GO-3 Preview","go-3 preview","go-3-preview","go3 preview","agibot go-3","agibot go-3 preview","智元 go-3"],"aliasNote":null,"origin":"tracked","tracked":{"trackedSince":"2026-10-29","announced":"2026-10-29","blurb":"AGIBOT's next-generation embodied foundation model for robot manipulation, shown at AGIBOT Innovation Day 2026. It is built on WRAM (World Reasoning Action Model), which links task reasoning, prediction of the scene's next state and action generation. Post-training mixes real robots, simulation and a world model, and deployed robots keep learning from their own successes and failures. No weights or API yet; AGIBOT says it will be open sourced by the end of 2026.","api":"none","openWeights":"announced","facts":[{"label":"Type","value":"Embodied foundation model (robot manipulation policy), not a chat model"},{"label":"Architecture","value":"WRAM: a VLM reasoning path plus CoAE (control-oriented autoencoder) visual encoding; SVP (single-step visual planner) predicts the future state and an inverse dynamics module turns it into actions; KASO keeps the candidate futures that fit the real actions during training"},{"label":"Training","value":"Pre-training on first-person human video and robot trajectories; post-training in Genie Evolver (about a hundred real robots, Genie Sim simulation, Genie Envisioner 2.0 world model); after deployment, the Genie Master agent framework turns successes and failures into memory and new training data"},{"label":"Predecessors","value":"AGIBOT's GO-1 (ViLLA architecture) and GO-2"},{"label":"Weights","value":"Not released; AGIBOT says GO-3 Preview and Genie Master will be fully open sourced by the end of 2026"},{"label":"API","value":"None announced"},{"label":"Parameters","value":"Not disclosed"}],"claims":["Lab claim: first in task success rate and process score on RoboDojo, highest overall score on RoboTwin and first in overall score on RoboColiseum, per results AGIBOT presented at the event.","Lab claim: CoAE compresses visual input to one sixteenth of DINOv3's tokens and cuts action reconstruction error by 35%.","Lab claim: on a real convenience-store deployment with Genie Master, full-task success rose from 28.6% on day one to 85.7% after five days, and single-task success was 17.7 points above standard fine-tuning (SFT)."],"sources":[{"kind":"lab","title":"AGIBOT press release: AGIBOT Introduces Worken at AGIBOT Innovation Day 2026","url":"https://www.agibot.com/article/231/detail/131.html","date":"2026-10-29"},{"kind":"lab","title":"AGIBOT Innovation Day 2026 post on X","url":"https://x.com/AGIBOTofficial/status/2115694835579666505","date":"2026-10-29"},{"kind":"press","title":"PingWest AI WAVE: AGIBOT's GO-3 Preview launch article (Chinese)","url":"https://www.pingwest.com/w/318950","date":"2026-10-29"},{"kind":"press","title":"Jiemian: GO-3 Preview and Genie Master to be open sourced by end of 2026 (Chinese)","url":"https://www.jiemian.com/article/15262681.html","date":"2026-10-29"}],"firstBenchmark":null},"organization":{"id":"agibot","name":"AGIBOT","slug":"agibot","country":null,"url":"https://www.agibot.com","color":"#11110F","mark":"A"},"price":null,"answer":{"word":"Pending benchmarks.","scope":"Announced Oct 29, 2026; no independent results yet.","lead":"As of Nov 3, 2026, GO-3 Preview is pending benchmarks on isitsota: announced Oct 29, 2026, with no independent results on the boards we track yet.","confidence":"low"},"status":"PENDING_BENCHMARKS","verdicts":[]}