The launch follows Prime Intellect's acquisition of Moondream, a startup described as a two man frontier lab with fewer than 10 employees. Founders Vikhyatk and Jason Allen have joined the company to lead inference development. Prime Intellect plans to integrate Photon into its broader computing stack to lower costs and attract developers building autonomous AI agents and robotics.
Wednesday, October 21, 2026
Story
Prime Intellect Launches Moondream AI on OpenRouter with 60 Inferences Per Second
AI5 days ago
Developer access to high-speed image analysis has expanded to a new API hub through a deployment by Prime Intellect. The Moondream 3.1 and Preview multimodal models are now available on OpenRouter, supporting real-time object detection and pointing within chat interfaces. These models run on Photon, an inference engine that compiles AI into single GPU programs, achieving more than 60 inferences per second on one Nvidia H100 with response times in the tens of milliseconds. In object-pointing tests, Moondream 3.1 reached 80% of the accuracy of Muse Wonder at roughly 1,000th of the cost.
Looking for more context or analysis?
Tilt chat provides reliable, tailored insights, alerts, and automations.