Simular is an autonomous agent company founded in 2023 by former DeepMind researchers. The team, which draws from DeepMind, Google, and other leading AI laboratories, builds AI agents capable of operating computers - browsers, desktops, and smartphones - in the way a human user would. The company has raised $21.5M in funding and counts over 12,600 users on its platform.
Its flagship product, Agent S3, achieves a 72.6% success rate on OSWorld (a computer-use benchmark), 90.1% on the WebVoyager browser benchmark, and 71.6% on AndroidWorld smartphone tasks. Underpinning these results is a neuro-symbolic framework for lifelong reinforcement learning, which combines large language models with symbolic code to balance flexibility and precision. Simular also publishes open-source agentic AI frameworks and contributes research to top academic conferences.
The company's technical work spans several domains:
- Agentic AI and autonomous computer operation
- Browser, desktop, and mobile automation
- Neuro-symbolic methods and lifelong reinforcement learning
- Large language models and open-source framework development
Simular operates at the intersection of applied AI research and product development, with an emphasis on translating advances in reinforcement learning and symbolic reasoning into systems that automate real-world computer workflows.