Model family

λ

Participating organizations

Version scope and roles for these associations await verification.

View evidence

λ₀ is a humanoid vision-language-action model from Alibaba Token Hub and Amap, Alibaba Group. It learns whole-body loco-manipulation from egocentric human data and predicts motion tokens and hand commands from images, instructions and proprioception. Authors report experiments on Unitree G1 with Revo2 hands.

Overview

Model type
Action policy
Listed versions
λ₀
Family introduction

Model history

October 2026

λ₀ technical report advances from v1 to v2

New resourcesOfficial announcement

arXiv records v2 on October21;the catalog cites this fixed revision for architecture and evaluation protocols. The revision date does not establish a new model version,a capability breakthrough or a weight/code-license change.

Previously reported

The arXiv report’s v1 was submitted on October13.

This announcement

arXiv records v2 on October21;the catalog cites this fixed revision for architecture and evaluation protocols. The revision date does not establish a new model version,a capability breakthrough or a weight/code-license change.

Ordinary history:an officially dated report revision without independent evidence of a material capability or availability change.

OpenLambda opens code and human-data alignment tools

Access changeOfficial announcement

Official news dates the openlambda code opening to October14,including training reference code,human collection,motion alignment and examples. The pinned code license is Apache-2.0. Current Stage-III configs are examples and complete evaluation/deployment remain TODO. Do not assign this date to first weight availability or claim full HumanVerse access.

Previously reported

The model and HumanVerse research had been introduced on October13;the next dated repository news records the code-opening step.

This announcement

Official news dates the openlambda code opening to October14,including training reference code,human collection,motion alignment and examples. The pinned code license is Apache-2.0. Current Stage-III configs are examples and complete evaluation/deployment remain TODO. Do not assign this date to first weight availability or claim full HumanVerse access.

Featured because public code and alignment tooling materially change developer access from research description to inspectable implementation. This is a code-access event,not proof of complete reproducibility,public full training data,hostedAPIrights or a ready-to-run robot policy.

λ₀ introduces whole-body human-data mid-training for humanoid control

New modelOfficial announcement

The October13 research introduction presents λ₀ and HumanVerse:500hours of synchronized human video/body/hand experience feeding three-stage learning. The event date follows arXivv1 submission and the official repository news. Model results,alignment-only progress and downloadable Stage-II weights have distinct scopes;this introduction is not proof of a task-ready public policy.

Previously reported

Humanoid demonstrations depend on robot operation,while hand-only egocentric supervision omits locomotion and posture.

This announcement

The October13 research introduction presents λ₀ and HumanVerse:500hours of synchronized human video/body/hand experience feeding three-stage learning. The event date follows arXivv1 submission and the official repository news. Model results,alignment-only progress and downloadable Stage-II weights have distinct scopes;this introduction is not proof of a task-ready public policy.

Featured for a named humanoid VLA and a materially different whole-body data/three-stage learning approach,not its admission score or an unqualifiedSOTA claim. Reported simulation uses task-wise configuration selection;real success and milestone progress differ. Full data and deployable evaluated weights are not established as public.

View all important updates →

Papers & reports (1)

λ₀: Scaling Open-World Egocentric Human Data for Humanoid Loco-Manipulation

Technical report

arXiv: 2610.18950DOI: 10.48550/arxiv.2610.18950
Authors & affiliations (16)

Author order: Team authorship

  1. Lambda-humanoid Team

    Alibaba Token Hub · Amap · Alibaba Group

  2. Chongyang Xucore contributor

    Not provided

  3. Jin Chencore contributor

    Not provided

  4. Jinhui Yecore contributor

    Not provided

  5. Yilun Chencore contributor; project lead

    Not provided

  6. Yiming Jiangcore contributor

    Not provided

  7. Zhao Wucore contributor

    Not provided

  8. Hu Weicontributor

    Not provided

  9. Lin Qucontributor

    Not provided

First published
2026-10-13
Latest paper revision
v2 · 2026-10-21
Venue
Not provided

Covered model versions

  • lambda · λ₀

    λ₀researchmodel andthree-stagepipeline;reportedStageIIIevaluationsareversion-owned,notattributedtothepublicrawStageIImidtraincheckpoint. SIMPLEtask-wiseconfigurationselection andpartial-progressscopesretained.

Paper ↗PDF ↗

Official resources

Submit a source or correction