Deep Research
#13 in KI-Söök & Researchgoogle · siet 21. April 2026 (Public Preview von Deep Research und Deep Research Max auf Basis von Gemini 3.1 Pro) · 16× · tolest 30. Juni 2026
Deep Research and Deep Research Max are autonomous AI research agents from Google DeepMind, offered via the Gemini API's Interactions API. They are built on the Gemini 3.1 Pro model, autonomously browse the open web plus private/proprietary data sources (including via Model Context Protocol), and produce multi-step, fully cited reports with native charts and infographics. Deep Research is optimized for low-latency interactive use cases, while Deep Research Max uses extended test-time compute for comprehensive, asynchronous research workflows requiring maximum analytical depth. Both agents are currently in public preview and are exclusively accessible through the Interactions API, not via generate_content.
Features
| Output Formats | Cited text reports in Markdown with inline generated HTML charts/infographics (also via Nano Banana) |
| Base Model | Gemini 3.1 Pro |
| Integrations | Google Search, URL Context, Code Execution, File Search, remote MCP servers (incl. FactSet, S&P, PitchBook in development) |
| Collaboration | Collaborative planning: users can review and adjust the research plan before execution (collaborative_planning parameter) |
| License | Proprietary cloud service under Gemini API Additional Terms; documentation content under CC BY 4.0, code samples under Apache 2.0 |
| Platform | Gemini API (Interactions API), Google AI Studio, soon Google Cloud/Vertex AI for enterprises |
| Price | No fixed pricing model; billed based on token usage of Gemini 3.1 Pro plus Google Search fees (paid API tier only) |
| Release Date | April 21, 2026 (public preview, successor to the December 2025 preview) |