---
type: "post"
id: "so-arm102-an-open-source-so-arm102-robot-bb4ac7"
slug: "so-arm102-an-open-source-so-arm102-robot-bb4ac7"
title: "SO-ARM102: An open-source SO-ARM102 robot arm, successor to SO-ARM100 and SO-ARM101"
description: "Robonine’s SO-ARM102 is an open-source leader–follower robot arm kit for teleoperation and imitation-learning demonstrations, succeeding the SO-ARM100 and SO-ARM101."
retrieval_nugget: "Robonine’s SO-ARM102 is an open-source leader–follower robot arm kit for teleoperation and imitation-learning demonstrations, succeeding the SO-ARM100 and SO-ARM101. Its repository pairs a LeRobot-compatible setup with build files, assembly guidance and stated performance comparisons, giving robotics teams a concrete reference for creating a small demonstration-capture rig."
published_at: "2026-11-17"
updated_at: "2026-11-17"
record_date: "2026-11-17"
date_kind: "published_at"
topics: ["open-source"]
entities: ["Robonine","SO-ARM102","SO-ARM101","SO-ARM100","LeRobot"]
related_patterns: []
novelty_label: "new_fact"
what_changed: "SO-ARM102 is described as an open-source, LeRobot-compatible successor to SO-ARM100 and SO-ARM101 for teleoperation and imitation learning."
prior_refs: []
source_url: "https://github.com/roboninecom/SO-ARM102"
source_urls: ["https://github.com/roboninecom/SO-ARM102"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:so-arm102-an-open-source-so-arm102-robot-bb4ac7"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/so-arm102-an-open-source-so-arm102-robot-bb4ac7/","markdown":"https://newruntime.com/posts/so-arm102-an-open-source-so-arm102-robot-bb4ac7.md","json":"https://newruntime.com/posts/so-arm102-an-open-source-so-arm102-robot-bb4ac7.json"}
markdown_url: "https://newruntime.com/posts/so-arm102-an-open-source-so-arm102-robot-bb4ac7.md"
---

# SO-ARM102: An open-source SO-ARM102 robot arm, successor to SO-ARM100 and SO-ARM101

## Retrieval answer

Robonine’s SO-ARM102 is an open-source leader–follower robot arm kit for teleoperation and imitation-learning demonstrations, succeeding the SO-ARM100 and SO-ARM101. Its repository pairs a LeRobot-compatible setup with build files, assembly guidance and stated performance comparisons, giving robotics teams a concrete reference for creating a small demonstration-capture rig.

## Summary
Robonine’s SO-ARM102 is an open-source leader–follower robot arm kit for teleoperation and imitation-learning demonstrations, succeeding the SO-ARM100 and SO-ARM101. Its repository pairs a LeRobot-compatible setup with build files, assembly guidance and stated performance comparisons, giving robotics teams a concrete reference for creating a small demonstration-capture rig.

## What happened
Robonine describes SO-ARM102 as a leader–follower kit: an operator moves the leader by hand and the follower repeats its motion. Each arm has five degrees of freedom plus a gripper. A single-arm configuration uses one leader and one follower; a bimanual setup places two kits side by side, with each hand controlling a leader. The project positions these configurations for teleoperation and for recording demonstrations for imitation learning, and says the software setup uses the SO-ARM 102 follower plugin with the standard `so101_leader` type in LeRobot.

The design succeeds SO-ARM101 with a stiffer PET-CF frame and a parallel gripper. Robonine’s comparison lists an 85 mm span for the two parallel jaws, versus an angular jaw on SO-ARM101. It reports 3.4–3.9 mm of frame deflection at 300 g, down from 7.5 mm; the test used printed servo dummies at joints 2, 3 and 4, with SO-ARM101 printed in PLA and SO-ARM102 in Fiberon PET-CF17 at 25% infill. The comparison also lists a rated payload increase from 150 g to 300 g, shoulder and elbow no-load speed rising from 45 to 75 RPM, and base rotation increasing from 220 to 352 degrees. Robonine says the speed figures are manufacturer no-load ratings at 12 V and the payloads are rated values.

The repository includes 3MF slicer projects, STL meshes, STEP sources, a bill of materials, printing instructions, an assembly guide and a 32-page booklet identified as version 0.8. Individual parts fit a 180 × 180 mm printer bed; the largest is 160 mm long. For one leader-and-follower set, the project estimates component costs at $342.40 and printing material at another $46, for $388.40 total before shipping and taxes. Buying full packs and filament from scratch is estimated at about $430, including unused spares. The source excludes tools, computer, printer, electricity and assembly labor, and warns that supplier prices, variants and minimum order quantities should be checked before ordering.

Robonine also publishes standalone URDF and Xacro exports, with mesh paths relative to the export folder. The repository describes its mass properties as estimates based on specified printing assumptions, and records geometry and inertia checks. Its licensing page distinguishes hardware designs under CERN-OHL-P-2.0, software and configuration under Apache-2.0, and documentation and images under CC BY 4.0.

## What the archive adds
We have not covered this before; it opens a new thread.

## Trend signal
This does not yet fit a pattern New Runtime tracks. The release concerns an open robot-arm design and demonstration-capture setup, rather than one of the tracked shifts in software-agent interfaces, runtimes or organizational practice.

## Why it matters
For teams exploring embodied AI, demonstration collection can be a practical bottleneck: a model needs examples of the actions and interactions it is expected to reproduce. A leader–follower rig gives operators a direct way to produce those demonstrations, while the repository’s CAD, print layouts, BOM and LeRobot setup make the hardware path inspectable rather than a black-box purchase. The published comparisons also make design trade-offs more legible: teams can examine the stated payload, reach, backlash and stiffness changes instead of relying only on a general claim of improvement.

This is not evidence that the arm will improve a particular policy or work reliably across laboratories. It is a reproducible starting point for a narrow class of manipulation experiments. Builders should budget for parts and iteration, validate the physical fit, and assess whether the arm’s range and payload match the objects and tasks they actually need to record.

## Limits and uncertainty
The performance comparisons are Robonine’s reported measurements and specifications; they are not independently verified here. The stated deflection test compares different frame materials and infill conditions, so it should not be read as a controlled isolation of every design change. Payload and servo-speed figures are identified as rated values, not demonstrations of sustained performance in a working task. Cost estimates are reference figures rather than supplier quotes, and exclude shipping, taxes, tools, equipment and labor. The project does not establish how well demonstrations recorded with this kit transfer to a particular robot, policy, object set or deployment environment.

## Skill to build now
If a team is considering a demonstration rig, turn the repository into a pilot checklist before buying: match the BOM’s servo variants and power supplies, confirm printer-bed dimensions, and review the v0.8 assembly booklet alongside the step-by-step guide. Robonine specifically warns not to swap the leader’s 5 V and follower’s 12 V supplies, and flags variable carbon-rod diameters and cable lengths as assembly concerns. Those checks are concrete ways to avoid preventable rework in a first build.

## Reader takeaway
SO-ARM102 is more than a robot-arm announcement: it is a documented, LeRobot-compatible route to teleoperation and imitation-learning demonstrations, with files and cost assumptions that teams can inspect. Treat its headline improvements as the project’s own reported figures, and treat the BOM as a planning estimate—not a guaranteed delivered cost or proof of task performance.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Open Source](https://newruntime.com/topics/open-source.md): Explore the open-source topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Sources](https://newruntime.com/sources/): See the original sources this archive cites.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)
