What it is
Manus describes its product as an agent that plans and executes work in a computer environment. Its documentation describes internet access, a persistent file system and software execution in a sandbox. These are vendor-described capabilities, not OpenModelsHQ test results.
Manus belongs in an agent-product comparison rather than the open-weight model catalogue. Its published context-engineering article explains design choices around tool use, file-based memory and keeping useful execution history. That article describes the team's approach at publication; it is not a complete specification of today's service.
Documented capabilities
- Task planning
- Browser and file work
- Sandbox execution
- Deliverable generation
These labels summarise the cited documentation. They are not an independent assessment of reliability or task success.
Questions to ask before choosing
- Can you inspect the sources and intermediate files behind a result?
- Which actions require your approval, and can you stop a task?
- What happens to uploaded files and connected accounts?
- How are usage, export options and plan limits calculated?
Understand the surrounding harness before treating a product comparison as a model comparison.