Simular AI

On-device / Computer-use AI agents that autonomously operate browser and mobile apps via reinforcement learning. Simular's research vision: 'AI shouldn't be built in closed labs -- but out in the wild.' Self-developed multi-modal policy model + RL framework. Simular tops benchmarks across browsers (WebVoyager 90.1%), computers (OSWorld 72.6%), and smartphones (AndroidWorld 71.6%) simultaneously — the only agent doing so. ICLR 2025 Best Paper at Agentic AI for science workshop.