Welcome to the Arm AI Optimization Challenge 2026. We’re inviting developers to build and submit projects that show how AI can be optimized for Arm-powered platforms across three challenge tracks:
For more details information for each Track, please visit the Track Details Tab.
Across all tracks, submissions should show clear optimization work and measurable improvements where possible. Optimizations we will look for include model size (reduce size on disk or in memory), model quality (improve fine-tuning or output quality), model speed (improve tokens/sec, time to first token), inference server speed (improve throughput, latency), developer experience (improve tools, workflows, setup, documentation), and Arm-specific optimization (implement optimizations in an existing framework, library, model, or application to run better on Arm). Developers can use Arm Performix to get exact benchmarks of their Arm based performance and be able to clearly show their results.
Requirements Submissions to the Hackathon must meet the following requirements: - Include a Project built with the required developer tools and meets the above Project Requirements. - Provide a URL to your code repository for judging and testing. The repository must contain all necessary source code, assets, and instructions required for the project to be functional. The repository must be public and open source by including an open source license file. This license should be detectable and visible at the top of the repository page (in the About section). MIT or Apache 2.0. - Include a text description that should explain the features and functionality of your Project. - Project Overview: A brief description of the project and its purpose. Also explain what makes it interesting and why it should win. - Functionality / Output: Explain what the project does and what the final output is (optimized model, migration example, scavenger deliverables, etc.). - Setup Instructions: Step-by-step instructions on how to build/run/validate on an Arm-powered device or Arm64 environment (as applicable to your track). - Optional: Include a demonstration video of your Project (less than three minutes, publicly visible on YouTube, Vimeo, or Youku, showing the Project functioning on the device). It must not include third party trademarks, or copyrighted music or other material unless the Entrant has permission to use such material. - Track 1 & Track 2: Each submission must include a copy of the project’s source code, either attached directly or linked to an open-source repository (e.g., GitHub). - Track 3: Each submission must include proof artifacts (links/screenshots) as described in the track requirements.
Prizes - Overall Winner: Project featured in the Arm Community Blog. - Overall Runner Up: Project featured in the Arm Community Blog. - Best in Category: Physical AI: Project featured in the Arm Community Blog. - Best in Category: Cloud AI: Project featured in the Arm Community Blog. - Best in Category: Mobile AI: Project featured in the Arm Community Blog.
Judging Criteria - Technological Implementation – 40 points: Does the submission demonstrate quality software development? Does it clearly leverage Arm-powered platforms (on-device, Arm64, efficiency-minded design)? Is the technical approach sound and well executed? - User Experience / Developer Experience – 15 points: Is it clear how to use, run, or validate the project? Is the documentation well structured? Could this be taken further or reused by other developers? - Potential Impact – 20 points: How useful is this to the developer community? Does it create reusable artifacts; optimized models, migration templates, prompt assets, or learning-ready content? - “WOW” factor – 25 points: How creative and compelling is the submission? Does it stand out in approach, usefulness, or clarity? Can it quickly capture attention and communicate value?