Arduino VENTUNO Q

Arduino VENTUNO Q is a single-board computer and developer board for edge AI, computer vision, and robotics. Arduino first unveiled it on March 9, 2026.^[3] Public pre-orders opened on August 25, 2026.^[4][5][11] Arduino S.r.l. is a subsidiary of Qualcomm Technologies.^[4]

The board combines a Qualcomm Dragonwing IQ-8275 application processor with an STMicroelectronics STM32H5F5 microcontroller. Arduino calls this a dual-brain design: Ubuntu and AI workloads run on the Qualcomm processor, while Arduino Core on Zephyr RTOS runs on the STM32 microcontroller for time-sensitive I/O and control. The two sides communicate through an RPC bridge.^[1][2][5]

Scroll sideways for more →

History and status

Arduino's March 2026 announcement presented VENTUNO Q as an upcoming board and said it would be available soon.^[3] It did not report that units were shipping. On August 25, Arduino, Qualcomm, and Canonical separately announced that public pre-orders had opened.^[4][5][11] Arduino also promoted the opening in an X post on that date.^[10]

The distinction between a product announcement, a pre-order, and a shipment matters. On August 27, the US Arduino Store listed the board at an introductory price of USD 299 and estimated delivery in four weeks.^[8] The regional Arduino Store displayed EUR 298.99 including VAT, also described the price as introductory, and showed a four-week delivery estimate, but its captured page showed the item as sold out.^[9] These were storefront snapshots, not a permanent global price or a guaranteed delivery date. Arduino and Qualcomm did not give a fixed worldwide general-availability date in the cited launch materials.^[1][4][8][9]

Arduino's August launch promotion offered a 65 W USB-C power supply and a 24-pin USB-C cable through September 30, 2026 while supplies lasted.^[11] That limited promotion should not be read as a permanent description of the standard package contents.

Dual-processor architecture

The main processor is a Qualcomm Dragonwing IQ-8275, part of the Dragonwing IQ8 series. It integrates an eight-core Kryo CPU, Adreno 623 graphics, a Spectra 692 image-signal processor, and a Hexagon neural-processing unit.^[1][2][6] Arduino lists the CPU at up to 2.36 GHz, while Qualcomm's IQ8 table lists the IQ-8275 at up to 2.35 GHz.^[1][6] The small difference is consistent with rounding in the published specifications; neither source identifies a separate VENTUNO Q processor version.

Arduino specifies the board's Hexagon NPU at up to 40 dense TOPS.^[1][2] Qualcomm's broader IQ-8275 documentation describes 20-TOPS and 40-TOPS configurations, and Arduino identifies VENTUNO Q with the 40-TOPS configuration.^[6] TOPS measures a maximum rate of arithmetic operations at a stated numeric precision. It is not a benchmark for a complete application, and it does not by itself establish model accuracy, token throughput, camera frame rate, or response latency.

The second processor is an STM32H5F5 microcontroller. Arduino and STMicroelectronics document an Arm Cortex-M33 core at 250 MHz, 4 MB of flash, and 1.5 MB of RAM.^[1][2][7] Arduino assigns real-time I/O, PWM, CAN-FD, and motor-control work to this side of the board. Its product page advertises response below 1 ms, but it does not provide an end-to-end test method or latency distribution.^[1] The figure is therefore a vendor specification, not a guarantee for every control loop or connected device.

The processor pair reflects the board's intended physical AI role. Linux applications can perform perception or inference on the IQ-8275, then call microcontroller functions through Arduino's RPC bridge. The STM32 side can continue handling timing-sensitive interfaces independently of Linux scheduling.^[1][2] This division is an architectural option, not proof that any particular robot will meet a safety or real-time certification requirement.

Memory, storage, and connectivity

VENTUNO Q has 16 GB of LPDDR5 memory, listed in the FAQ as two 8 GB devices, shared by the main CPU, GPU, and NPU. It has 64 GB of soldered eMMC storage and an M.2 connector for an NVMe Gen4 drive.^[1] The M.2 slot provides expansion, but Arduino does not include an SSD capacity in the base-board specification.

The board's documented networking includes tri-band Wi-Fi 6 on 2.4, 5, and 6 GHz, Bluetooth 5.3, and a 2.5 Gbit/s Ethernet port. Display paths include HDMI, MIPI DSI through a carrier header, and DisplayPort Alt Mode over USB-C. Camera interfaces include USB camera support and three MIPI CSI connectors, with two CSI paths also exposed through the JMEDIA header.^[1][2]

For control and expansion, Arduino documents one CAN-FD interface with a physical transceiver on a screw terminal, three additional CAN-FD interfaces without transceivers on the JOMEGA header, and another no-transceiver CAN-FD connection on the UNO headers.^[1] It also exposes PWM, GPIO, Qwiic, UNO-style shield headers, a 40-pin Raspberry Pi-style header, and JMEDIA, JMISC, and JOMEGA carrier connectors.^[1][2] The presence of a familiar header does not make every accessory electrically interchangeable; voltage levels, pin assignments, power limits, drivers, and mechanical clearance still need to be checked for each design.

Software and model support

The IQ-8275 side ships with Ubuntu pre-loaded, and Arduino says the board includes an Ubuntu Pro license. Standard Linux development tools can be used instead of Arduino's own environment.^[1][5] Arduino App Lab provides a combined workflow for Arduino sketches, Python programs, and containerized AI components. On the microcontroller side, the Arduino Core runs on Zephyr RTOS.^[1][2][5] Current Arduino material says Debian support is coming soon, so Debian should not be treated as a shipping operating-system option.^[1][2]

Arduino's current FAQ lists support for the following model families and tasks:^[1]

• Qwen 3 4B as a local large language model, plus Qwen 2.5 7B and Qwen 3 4B vision-language models
• Gemma 4 E2B and E4B
• Whisper automatic speech recognition
• Melo and Piper text-to-speech
• YOLOX Small object detection
• MediaPipe gesture recognition

The FAQ also mentions llama.cpp with GGUF models, Qualcomm's GenieX runtime, Qualcomm AI Hub models, PyTorch, and custom-model workflows through Edge Impulse.^[1] These statements describe software compatibility and supported workflows. They do not supply comparable throughput, memory-use, accuracy, context-length, or energy measurements for the listed models. Model availability and optimized files can also change independently of the board hardware.

Arduino and Canonical describe VENTUNO Q as compatible with ROS 2.^[1][5] The board's combination of camera input, local inference, Ethernet, CAN-FD, GPIO, and a dedicated microcontroller makes it a development platform for perception and actuation experiments. ROS 2 compatibility does not mean every ROS package is hardware-accelerated or that the board is certified for safety-critical control.

Evidence and limitations

Most public evidence about VENTUNO Q comes from Arduino and its platform partners Qualcomm, Canonical, and STMicroelectronics. As of August 27, 2026, the cited materials did not include an independent benchmark of a retail board. The 40 dense TOPS value is an upper-bound accelerator specification, and the under-1-ms control value is an Arduino claim without a published test protocol.^[1][6]

The product page also uses broad phrases about local models, accessory compatibility, and deployment readiness. Those phrases should be read alongside the detailed specifications. A workable model configuration depends on quantization, runtime, memory allocation, prompt or image size, thermal behavior, and software versions. An accessory depends on its electrical and driver requirements. A prototype intended for production, industrial control, or autonomous operation needs separate validation for reliability, security, real-time behavior, and any applicable safety standard.

The board was still a pre-order product at the research cutoff. Pricing, stock, the four-week estimate, the promotional bundle, Debian support, model lists, and software images were all subject to change.^[1][8][9] The stable facts were the announced processor pair, memory and storage configuration, documented connectors, and the split between Ubuntu on the IQ-8275 and Arduino Core on Zephyr RTOS on the STM32H5F5.^[1][2][4][5]

References

1. ^Arduino. "Arduino VENTUNO Q - Where AI Takes Action." Accessed August 27, 2026. arduino.cc/product-ventuno-q
2. ^Arduino Documentation. "VENTUNO Q." Accessed August 27, 2026. docs.arduino.cc/...ventuno-q
3. ^Arduino Team. "Introducing Arduino VENTUNO Q: Your New AI, Robotics, and Actuation Platform!" Arduino Blog, March 9, 2026. blog.arduino.cc/...robotics-and-actuation-platform
4. ^Qualcomm. "Arduino VENTUNO Q Is Here: Bringing Agentic AI to the Physical World with Qualcomm Dragonwing IQ8 Series." August 25, 2026. qualcomm.com/...nging-agentic-ai-to-the-physical-w
5. ^Canonical. "Arduino VENTUNO Q Is Available for Pre-Order with Ubuntu Pre-Installed." August 25, 2026. canonical.com/...e-order-with-ubuntu-pre-installed
6. ^Qualcomm. "Qualcomm Dragonwing IQ8 Series Platform Product Brief." Revision A, 2025. docs.qualcomm.com/...ies_Product_Brief________.pdf
7. ^STMicroelectronics. "STM32H5 Series." Accessed August 27, 2026. st.com/...stm32h5-series
8. ^Arduino US Store. "Arduino VENTUNO Q." Accessed August 27, 2026. store-usa.arduino.cc/...ventuno-q
9. ^Arduino Store. "Arduino VENTUNO Q." Accessed August 27, 2026. store.arduino.cc/...ventuno-q
10. ^Arduino. "The Wait Is Over! Meet Arduino VENTUNO Q, Where AI Takes Action." X, August 25, 2026. x.com/...2092253876137205982
11. ^Arduino Team. "The Age of Physical AI Is Here, and It Runs on the Arduino VENTUNO Q Board." Arduino Blog, August 25, 2026. blog.arduino.cc/...-on-the-arduino-ventuno-q-board