AI got smarter. Its computer didn’t.
AI today works through computers designed for people: a screen to look at, a mouse to point with. At each step they look at a picture of the screen, work out what is on it, act, and look again. Models have grown far more capable, but they still spend much of that capability working around a machine built for someone else.
Pine Computer starts from the model instead. It reads a web page as structure (what is on it, what can be done with it, what just changed) rather than reconstructing it from a picture each time. Pictures stay for people, who can watch the live screen, take the controls and hand them back.
A computer a product can put to work
For developers, the entry point is the SDK. A product hands Pine Computer a task, follows its progress, and receives the result: the files, records and answers the task produced. Pine runs the computer, the intelligence in it, the browser and the isolation; the developer builds the experience around it.
Each Pine Computer runs sealed in its own sandbox, and the developer’s keys stay with them. A live screen can sit inside the product, so its users can step in when a task needs them. Developers who want to bring their own model can connect it too; it works from screenshots today.
Pine AI uses Pine Computer in its own assistant and enterprise work. In one enterprise deployment it helps automate audits, for a customer whose team now takes on 50% more work with the same people.
Andrew Mackenzie, co-founder of Subliminal, which evaluated Pine Computer for its product, said Pine had “thought about all of this and built it all in.”
The results
In Pine’s tests, Pine Computer is 2–5× faster than AI on conventional computers.[1] On UniPat AI’s SaaS-Bench v1.1, a public benchmark of 106 business workflows across 23 applications, it has the highest checkpoint score in the publisher’s table: 78.3%, against 74.3% for Opus 5 with Claude Code and 71.1% for GPT-5.6 Sol with Codex. It ran GPT-5.6 Luna, a lower-cost model, at about $1.02 per task in model cost, against $26.50 and $20.50 for those systems. A checkpoint score is the share of a task’s checkpoints passed, not tasks finished: Pine Computer completed fewer whole tasks than those systems (27.4%, against 31.1% and 29.2%), and these are comparisons between whole systems, each with its own software and budget, not a test of the computer alone. The evaluation notes give the definitions, and the run data is published. For more on Pine Computer’s architecture, capabilities and benchmark results, read the launch blog.
“We spent years making the model smarter. Now, we’re making the computer worthy of the model,” said Dylan Wang, Pine AI’s co-founder and chief architect.
Wang said Pine plans to open its designs, implementations and specifications, and to work with others on what a computer built for AI should be. “We don’t believe the future of computing for AI should belong to one company,” he said.
