femtoAI opens developer community for its compressed AI chip platform

What developers get access to

The San Bruno, California company, which describes itself as a leading full-stack platform for AI chips and compression, announced the move on 6 October 2026. Developers can now use the femtoAI Sparse Processing Unit (SPU) chip evaluation kit, compressed models, a compiler and a tool called Sparsity Studio. According to femtoAI, these components are meant to simplify the development of new AI-powered products.

The company says its approach turns patented sparsity methods for neural networks and advanced model compression into real savings. The result, according to femtoAI, is ten times lower power consumption and ten times lower memory requirements, with no loss in accuracy. Developers can explore the tools and systems at developer.femto.ai.

The main features of the community, as listed by the company:

• SPU evaluation kit: developers can order their own kit to get access to the SPU chip.
• Self-service development: access to the full range of femtoAI tools, the compiler and models.
• Model flexibility: bring your own model, use one of femtoAI's pre-built models, or build one from scratch, supported by documentation.
• Sparsity Studio: a measurement tool that shows the actual power and memory savings achievable with femtoAI's sparsity.
• Community forum: an ongoing open discussion and Q&A with femtoAI experts and other developers.

Customers and technical claims

Bill Hoppin, Chief Business Officer at femtoAI, said the company has been working behind the scenes with dozens of companies in the developer community. He said now is the right time to make the opportunity available to more companies and developers. According to Hoppin, customers such as Legato and Marshall gained a competitive advantage through easy development and deployment with the company's tools and systems.

Sangnam Choi, VP of Audio & Sensing at femtoAI, said the company applies advanced model compression techniques that have allowed it to deliver open-source models such as OpenAI Whisper with ten times lower memory requirements. The team has also built custom solutions such as local voice commands, he said, and continues to look for new techniques that further reduce power and memory needs.

The company says the platform has already contributed significantly to its audio and sensing work. Customers named in the release include smart glasses start-up Legato, headphone and earbud brand Marshall, and NewSound, which used the technology to build AI audio solutions that, according to femtoAI, use ten times less memory and ten times less power.

Mehul Trivedi, CEO of Legato Hearing, said the combination of easy-to-use tools, innovative features and strong support helped Legato reach a level of hearing performance for its hearing aids that was previously impossible in this form factor, in a remarkably short time.

Events and what it means

femtoAI is also hosting a developer happy hour as part of SF Tech Week on Wednesday, 7 October, at the Barebottle Brewing Company beer garden in Salesforce Park. The company promises drinks, small bites and an exclusive walkthrough of the developer community with femtoAI experts.

The company was founded in 2018 by researchers from Stanford University's "Brains in Silicon Lab" and is headquartered in Silicon Valley. It says its platform serves products ranging from wearables and household appliances to smart glasses and robotics. With the community now open, teams in those areas can order an evaluation kit and test the claimed savings with their own or pre-built models. The efficiency figures are the company's own claims, and Sparsity Studio is the tool intended to let developers measure the savings themselves.