Opera Turns the Browser Into a Research Machine

Deep-research agents are starting to reshape how people gather information online. Instead of producing single answers from a chat prompt, these systems divide a question into smaller pieces, search multiple sources at once, and merge the findings into a coherent result. They’re now becoming a standard feature inside emerging “agentic browsers,” which are built to automate complex web research rather than just display pages.

Into this context steps Opera Neon, the AI-driven browser from Opera Limited, which just announced its new deep research agent, Opera Deep Research Agent (ODRA). According to Opera, ODRA is the fourth specialised AI agent in Neon, joining its Chat, Make and Do agents. Beyond the usual, the browser is now designed to act as a platform for research workflows.

A browser built for research

Opera describes ODRA as model-agnostic and tool-agnostic, stating it “could combine the strengths of different models, like Gemini and GPTs,” and that it uses server-side parallelisation to “divide the problem into smaller ones and run separate researchers on them.” The implication is that rather than relying on a single large language model, Neon’s engine distributes subtasks across multiple models and synthesises the results.

In the press release, Opera states that ODRA has become “one of the very best deep research agents” on its internal DeepResearch Bench test, although details of the benchmark metrics are not publicly disclosed. The design reflects emerging academic work in multi-agent retrieval and generation, for example, the MA-RAG (“Multi-Agent Retrieval-Augmented Generation”) study shows that using multiple agents improves factual accuracy.