AlohaMini2, a dual-arm mobile robot you can print on a consumer printer

What changed between the first generation and the second

AlohaMini2 is a dual-arm mobile robot with a motorized vertical lift, described as a major generational upgrade over AlohaMini1 while keeping the self-build bill of materials under 1,000 dollars. The claims in the README are specific enough to be worth repeating, because a robot platform is judged on exactly these numbers.

The base max load goes from 10 kg to 70 kg. The lift capacity goes from 5 kg to 30 kg, and the lift is described as significantly reinforced. The arm payload goes from 0.3 kg to 1 kg, and the arm reach from 40 cm to 52 cm. The arm degrees of freedom go from 5+1 to 6+1. The camera count stays at five, and the README specifies the positions as forward, backward, chest and dual wrist.

The hardware changed in two structural ways. The arm system moves from SO-ARM100 and SO-ARM101 to AM-ARM200, a separate repository, and the mobile base system moves from mobile_base1 to mobile_base2. The printer requirement also changes, from a large-format FDM machine with a bed of at least 380 mm to a Bambu P2S, which the README describes as consumer-grade. That last change is the one that decides who can build this. A large format printer is a serious purchase for most labs, and a consumer Bambu machine is a purchase many of them already made.

The project states that AlohaMini2 is LeRobot-ready out of the box and remains fully open source, with hardware CAD, URDF and software included. The license is Apache-2.0, the last push was on 2026-07-01, and the repository has no GitHub releases.

The bill of materials, and what the 950 dollars actually buys

The bill of materials is given as a table of major components, and the README notes that it excludes the leader arm.

The servos come in two kinds. Eleven standard servos, the Feetech STS-3215 at 12V, for the follower arms and the base, at about 16 dollars each. Seven high-torque servos, the Feetech STS-3095 at 12V and 95 kg·cm, for the lift and the arm power joints, at about 50 dollars each. Two Waveshare bus servo controllers at about 5 dollars.

The compute is a Raspberry Pi 5 with 2 GB at about 93 dollars, and the README notes that compute can be swapped for a Jetson Orin Nano or any SBC with USB 3.0, which is a useful degree of freedom given that a Pi 5 is the minimum for the vision and the policy inference rather than a comfortable machine for it.

The perception is five 720p USB cameras with a 2.4 mm focal length at about 17 dollars each. The drive is three 127 mm omni wheels, 74A, at about 40 dollars. Power is two 12V 11200 mAh Li-ion batteries at about 16 dollars. The frame is 3D printed on a Bambu P2S, and the cost is expressed as about 6 kg of filament rather than a price. The total is about 950 dollars.

That number is worth interrogating, because it is a self-build cost rather than a purchase price. It counts components, filament and a Pi, and it does not count the printer, the tools, the shipping, the failed prints, or the engineering time. The README also points out that the 6 kg of filament is the printed structural parts alone, and the Quick Start puts printing all structural parts as a step in its own right.

The full bill of materials with CN and US buy links is at AlohaMini2/docs/BOM.md, which is where you would go before ordering anything.

Building it, one documented step at a time

The Quick Start is a four-step sequence, and each step has its own document in the repository.

The first step is the BOM, where you review the component list and source the parts. The second is 3D printing all the structural parts, about 6 kg of filament on a Bambu P2S, with the print guide at AlohaMini2/docs/print_guide.md. The third is assembly, described as about 120 minutes with the arms pre-assembled, with the guide at AlohaMini2/docs/assembly_guide.md. The fourth is software setup and teleoperation, with the guide in the software directory.

The project summary line says it can be assembled at home in about 120 minutes and that you train or deploy with LeRobot. The 120 minute figure applies to assembly only, and the arms arrive pre-assembled, so the honest total is that figure plus the print time plus the time to source and wait for the components.

The documentation set is one of the more complete parts of this project. Beyond the BOM, the print guide and the assembly guide, there is a software directory, and the repository has separate AlohaMini1 and AlohaMini2 directories, so the two generations sit side by side rather than the older one being deleted. That matters for anyone who built an AlohaMini1 and wants the simulation assets or the CAD for their existing machine.

The update history also shows what is published where. On 2025-12-18 the URDF and simulation assets became available under AlohaMini1/simulation/src/Aloha/, and on 2025-11-27 the STEP CAD files became available under AlohaMini1/hardware/mobile_base/step/. On 2026-02-26 a fine-tuning and deployment guide for Pi 0.5 on AlohaMini was added under examples/pi0.5_openpi, and on 2026-06-06 AlohaMini2 was released.

The point of a dated asset drop is that the assets exist in a specific place for a specific generation, and the dates tell you which generation each belongs to.

Two product variants and a hardware store

The product line table is short and it draws a distinction that matters for a research buyer.

AlohaMini2 uses STS-3215 standard servos on a 3D printed base, and it is aimed at self-build research and development. AlohaMini2 Pro uses STS-3250 industrial-grade servos on a metal-reinforced base, and it is aimed at higher-duty research and lab use. Both link to the same official store listing on Taobao, with the Chinese market treated as the primary channel for the hardware.

That Pro variant is the answer to the obvious objection to a 3D printed chassis with hobby servos, and it is worth noting that it exists at all, because a self-build kit with a Pro option that swaps the servos and reinforces the base is a different proposition from a hobby project. It is also not described in the performance tables, which are about AlohaMini2, so anyone evaluating the Pro needs to ask about its numbers.

The contact details are the other thing the table implies. There is an email address, a WeChat contact, and videos and tutorials on Bilibili, YouTube and TikTok. For a project with a Chinese hardware supply chain and a Chinese primary market, that set of channels is not unusual, but it does mean the assembly videos, which are where most of the practical knowledge of a self-build robot lives, are likely to be on Bilibili and in Chinese.

The acknowledgements name the lineage the design builds on, and they are the right starting point for understanding the architecture: ALOHA, LeKiwi, SO-ARM100, SO-ARM100-Track-Axis, Pi-0.5, LeRobot and Hugging Face. The ALOHA name is the origin of the bimanual low-cost approach, LeRobot is the framework the software targets, and Pi-0.5 is the policy model the example directory covers.

The project is created by Li Yiteng and Wu Zhiyong.

What this is for, and the limits of a printed platform

AlohaMini2 sits in a specific position: it is a research platform for embodied AI, and the deciding constraint is that a lab can afford several of them. At a 950 dollar self-build cost and a 70 kg base load, the economics work for a group that wants to run many experiments, and a mobile bimanual platform is something that is usually out of reach for anyone not working at a large laboratory.

The limit of the 3D printed approach is the same one that makes it valuable. Printed structural parts are good for a research chassis and are not good for a machine that will be dropped, loaded repeatedly to its rating, or run for years. The README is specific about the base load and the lift capacity, and those numbers are meaningful, but they are meaningful for a printed base in a lab, and the Pro variant with a metal-reinforced base and industrial-grade servos is the answer to the durability question.

The second limit is the compute. A Raspberry Pi 5 with 2 GB of RAM is the documented configuration, and the README explicitly says you can substitute a Jetson Orin Nano or any SBC with USB 3.0. That substitution is not a detail. A dual-arm system with five cameras running a vision policy in real time raises the question of whether a 2 GB Pi merely handles that workload or runs it comfortably, and the software is LeRobot-ready rather than LeRobot-optimised for a particular board.

The third limit is that this is a self-build. The 120 minute assembly is the easy part, because the arms come pre-assembled. The real work is the print job, the parts sourcing across two markets, and the debugging that follows the first power-on. The value of the repository is that all of that is documented rather than tribal knowledge, which is the difference between a project that is reproducible and one that is merely published.

The open CAD, the published URDF and the simulation assets are what turn it from a shopping list into a platform. A researcher can load the same robot into simulation, train a policy there, and then know that the physical machine matches the model, which is the condition that makes simulation work transfer at all.

Editorial conclusion

AlohaMini2 suits a research group that wants a real bimanual mobile platform without a four figure industrial purchase, and it suits someone who already works in LeRobot and wants hardware whose CAD and URDF are published rather than reverse engineered. It is the wrong choice if you need a turnkey robot, because the assembly time of about 120 minutes and the print job are on you, and it is the wrong choice if you need industrial repeatability, because the base is 3D printed and the servos are hobby grade Feetech units. Before starting, read AlohaMini2/docs/BOM.md for the full parts list with links, and check the ALOHA and SO-ARM100 lineage the acknowledgements name, because the arm system changed from SO-ARM100 and SO-ARM101 to AM-ARM200 between the two generations.

Frequently asked questions

Official sources

1. Issues
2. License: Apache-2.0
3. liyiteng/AlohaMini on GitHub
4. README