Measure before claiming
Performance depends on the complete configuration. We prefer published hardware, memory, latency, throughput, and quality results over broad superlatives.
[ˈneɪ.wə]noun
From Muaneywa. A thing you hang out to dry that never dries.
A symbol of tenacity.
More power to you
Intelligence belongs closer to people.
Neywa Labs builds the infrastructure that makes frontier open models run natively on personal devices.
Our work spans language, vision, speech, video, model optimization, inference, and native applications.
We develop open-source engines for builders and products for people who want capable AI without depending on the cloud for every interaction.
Modern AI is extraordinarily capable, but its intelligence is usually somewhere else.
Every request travels to a datacenter. Private context leaves the device. Applications depend on connectivity, external providers, variable latency, usage limits, and recurring inference costs.
That architecture is valuable, but it should not be the only option.
Apple devices have become powerful enough to run sophisticated language and multimodal models. When intelligence moves onto these devices, new experiences become possible: applications that respond immediately, work offline, understand local context, and keep sensitive information close to the user.
Neywa exists to make those experiences practical.
Local AI is a systems problem.
A model can be capable and still be unusable on a device for a myriad of reasons.
Useful local AI requires the model, runtime, hardware, and product experience to work as one system.
That is why Neywa works across the stack.
Trusted by more than 7M developers
We build in the open.
Open-source development is central to how Neywa works.
It lets developers inspect the implementation, test new architectures, contribute improvements, and build products on foundations they can understand.
It also creates a direct connection between research and application. When a new model is released, the work required to bring it on-device becomes reusable infrastructure for the wider ecosystem.
VLM
Inference and fine-tuning of Vision Language Models on your Mac
5.7k
5.1m- 5000+models
Audio
Speech, transcription and generation models running fully offline
- 8.1k
- 2.5m
- 5000+models
Our principles
Products complete the research
Infrastructure proves that something can work. A product proves that it matters.
The cloud should be a choice
Some workloads belong in the cloud. Others are faster, safer, more reliable, or more economical on the device. Developers and users should be able to choose.
What we are building toward
We believe local AI will become an important part of the computing stack.
Models will need to be matched to devices and use cases. Applications will move across different hardware configurations.
Developers will need to understand expected memory, quality, latency, throughput, and power requirements before they ship.
Neywa is building the technical foundation for that future - beginning with the Engine and Nativ.
Meet the founders
Engineers, visionaries and builders at heart.
Prince Canuma
Co-founder, CEO
Prince spent years building in open source an ecosystem enabling builders across the world to run powerful, state-of-the-art AI models on the hardware they already own. His mission is to give hundreds of millions access to the best possible Intelligence, irrespective of where they come from.
Sebastien Michaud
Co-founder, COO
Seb's career is shaped by commercializing unreleased solutions to critical problems. His mission at Neywa is to remove the tradeoff between access and agency every AI user currently faces: an ecosystem which distributes Intelligence with no data trade, no tax on growth, no cap on scale.
Lucas Newman
Co-founder, CTO / Head of Research
Lucas' career is marked by solving hard engineering problems that actually bring zero to one innovations to market. His mission is to constantly push forward research on what is possible in Local AI, and ensure builders like him have access to the cutting edge within the Neywa Labs ecosystem.
Join Neywa
Work on the infrastructure and products defining how AI runs on personal devices.
You might implement support for a newly released multimodal architecture, reduce the memory required for a long-context model, improve real-time voice inference, build a native product experience, or design the tools developers use to understand model-device performance.
The work is difficult, visible, and close to the frontier.
You already own a supercomputer
Run it locally, in the open