Arm’s ‘Arm AI Portal,’ announced earlier this month, is drawing fresh attention as a hub connecting more than 22 million developers and their AI agents to pre-optimized models, according to Arm’s newsroom and trade press.

The portal offers models including Alibaba’s Qwen, Google’s Gemma and Ultralytics YOLO, along with performance and accuracy data, latency-memory-size comparisons, code examples and deployment workflows, Arm’s newsroom and finance outlet BigGo reported.

The platform uses the Model Context Protocol (MCP), which lets AI agents discover models, optimization data and deployment information on their own. Agent-ready resources were offered first through early access, and developers will soon be able to bring their own or proprietary models for optimization, per WiseVoter and Jon Peddie Research.

Benchmarks cited by Arm include Qwen3-TTS running four times faster on the Vivo X300, and YOLO26n showing gains of more than 40 percent on the Vivo X300 and Raspberry Pi 5, WiseVoter and Omega Technology Solutions Group reported.

The portal positions Arm at the center of the shift toward running AI directly on phones and edge devices.