Nexa SDK runs any model on any device, across any backend locally—text, vision, audio, speech, or image generation—on NPU, GPU, or CPU. It supports Qualcomm, Intel, AMD and Apple NPUs, GGUF, Apple MLX, and the latest SOTA models (Gemma3n, PaddleOCR).
Free
Launch Team / Built With
Render Sandboxes: a computer for every agentIsolated compute, persistent state and a real filesystem
Promoted
what kind of hardware is mos suited for this? Macbook M4 128GB ram or?
@zack_learner Awesome. How much ram is needed? I guess that's the bottleneck for local LLM?
@zack_learner @sentry_co you can run a 3B model with 16GB RAM!
@zack_learner @alanzhuly Right. VSCode and Chrome already eating most of that already 😅 I need a new macbook for this!
Very excited to see this! It also supports mobile app such as Nexa Studio!
@ren_zhang1 Haha sure
@ren_zhang1 Thanks Ryan! Mobile AI is the new trend.
And Linux...?
Nexa SDK is an impressive and versatile software development kit that significantly simplifies integration and accelerates app development. Its well-documented APIs and intuitive design make it accessible for both beginners and experienced developers. The SDK’s robust features, seamless performance, and reliability stand out, enabling quick implementation without compromising quality.
@sritama_bose Thanks for your warm words!
@sritama_bose Thanks for the support and we look forward to hearing your feedback.
Hi Alex! Nexa SDK sounds like a game-changer for developers dealing with cloud constraints and privacy concerns. The ability to efficiently run multimodal AI on-device without sacrificing speed and privacy is impressive. Congrats on the GitHub success – excited to see how the community grows here! 🚀
@alex_cloudstar Thanks Alex for the support. Would love for you to join our community: https://discord.com/invite/nexa-ai
As someone running models on my Mac, I love seeing local-first done right—Apple MLX support plus solid on-device performance is exactly what I want. Lower latency, better privacy, no surprise cloud bills. This makes iterating on ideas way faster.
@wenwen_cao MLX should be the go-to inference framework for Apple devices!
With Nexa SDK, developers get a unified runtime for models on CPUs, GPUs, and NPUs
@issa_keita that's true, thanks
@issa_keita That's right and also cross platform on compute, mobile, automotive, IoT, and XR.