BeagleY-AI
BeagleBoard.org FoundationOpen single-board computer from the BeagleBoard.org Foundation built around the Texas Instruments AM67A vision processor. It pairs a quad-core Cortex-A53 at 1.4 GHz with a dual C7x DSP and matrix multiply accelerator rated up to 4 TOPS, an IMG BXS GPU and Cortex-R5 real-time cores, with 4GB LPDDR4 and a 40-pin Raspberry Pi-compatible header. It targets vision, robotics and smart cameras.
OSHWA-certified with full board design files and a public datasheet; the AM67A silicon and its firmware blobs are the caveat the openness axis weighs. Verified 2026-08-13 via the OSHWA certification record, the beagley-ai repository, the design documentation and the TI product page.
Openness
5 high confidence- schematics
- open (schematics + 3D/mechanical files, CC-BY/CC-BY-SA-4.0, OSHWA-certified US002616)
- toolchain
- open (open Linux/U-Boot BSP)
- datasheets
- public (public TI AM67A datasheet)
- blobs
- required (TI DSP/WiFi firmware blobs a caveat)
Full board design files published under Creative Commons and OSHWA-certified, with an open toolchain and a public datasheet, though proprietary AM67A silicon and firmware blobs are a caveat. The OSHWA certification is US002616, recording hardware under CC-BY-4.0, software under the GPL and documentation under CC-BY-SA-4.0. The design page itself no longer prints the OSHWA UID, though the certificate and the repository README both still do.
- https://certification.oshwa.org/us002616.html recorded 2026-08-13
OSHWA UID US002616 for BeagleY-AI by the BeagleBoard.org Foundation, certified March 27 2024, licences recorded as Hardware CC-BY-4.0, Software GPL, Documentation CC-BY-SA-4.0
- https://github.com/beagleboard/beagley-ai recorded 2026-08-13
repository under a CC-BY-4.0 license with design and 3D directories and a link to https://certification.oshwa.org/us002616.html
- https://docs.beagleboard.org/boards/beagley/ai/03-design.html recorded 2026-08-13
"This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License" and "Design materials and license can be found in the BeagleY-AI git repository"; the board carries 2x C7x DSPs with MMA and a Beagle BM3301 wireless module based on the TI CC3301, the two firmware-bearing subsystems behind the recorded blob caveat
- https://www.ti.com/product/AM67A recorded 2026-08-13
"AM67x Processors datasheet (Rev. B)" published openly on the part page, status ACTIVE
Adoption
4 medium confidenceIn stock at $70.00 with volume pricing at Seeed and orderable via TI, backed by the established BeagleBoard.org community, though recent (2024) so installed base trails Raspberry Pi. The AM67A it is built on is ACTIVE at TI with evaluation boards in stock.
- https://www.seeedstudio.com/BeagleYr-AI-beagleboard-orgr-4-TOPS-AI-Acceleration-powered-by-TI-AM67A.html recorded 2026-08-13
$70.00, "In stock"
Capability
3 high confidence4 TOPS plus an IMG BXS GPU suits edge vision and quantized CNNs, but fixed 4GB RAM and modest TOPS rule out large LLMs. The board page reads "Dual general-purpose C7x DSP with Matrix Multiply Accelerator (MMA) capable of 4 TOPs" and 4GB LPDDR4, which sits below the 6 TOPS RK3588 boards and well below the Jetson Orin line.
- https://www.beagleboard.org/boards/beagley-ai recorded 2026-08-13
"Dual general-purpose C7x DSP with Matrix Multiply Accelerator (MMA) capable of 4 TOPs"; 4GB LPDDR4; Texas Instruments AM67A Arm-based vision processor
Verified 2026-08-13