August 28, 2026 · Research
Sai tops OSWorld 2.0, using least tokens
Beats GPT-5.6 and Claude Opus on a 108-task benchmark that takes skilled humans over an hour.

About Simular
If you’re reading this, you’re probably like us – a knowledge worker who spends well over 1,500 hours a year just clicking, typing and moving on the computer.
We're a research-driven company chasing one goal: full machine autonomy. Every product we ship moves us closer to a world where AI takes over digital labor, freeing people to develop and flourish.
Real life
Continual learning in real life is where actions produce consequences that feed back in.
Full-stack
We build not only the user-facing agent interface, but also the virtual machine that the agent runs on to free humans from devices.
Neurosymbolic
Our neurosymbolic framework combines the exploratory power of LLMs with the deterministic reliability of code, balancing capability and reliability.
We’re a small group of scientists and engineers from DeepMind, Google, and top AI labs. We’ve worked on cutting-edge reinforcement learning systems, compiler design, and real-world robotics — and now we’re applying that experience to build agents that actually work.
Built by team of AI experts from












Try what we've built — or help us build what comes next.