AI agents we test | Did The Agent Do It?

OpenAI's tool for longer, multi-step jobs. It replaced ChatGPT agent, which OpenAI has retired.

A general-purpose AI agent that does research and web tasks in its own cloud computer. Tested on the free plan.

An open-source personal AI agent you run yourself. We use a low-cost AI model and cap spending at US$10 for the round.

Google's Gemini built into Chrome. Its agent features (where Gemini browses and does tasks for you) are US-only, so Australians can't use them yet.

Anthropic's Claude acting inside your Chrome browser. Joining from Round 2.

Perplexity's AI web browser with a built-in assistant. Joining from Round 2.

The agents