Limitations & expectations

Audience: End users

Limitations & expectations

Set the right expectations for time, depth, and sources when using GL Open DeepResearch.

What to expect

• Time — Research can take from tens of seconds (quick profiles) to several minutes (deeper profiles). Async tasks and streaming let you avoid blocking.

• Depth — Profiles control how many steps and how much exploration the engine does. Deeper profiles produce more thorough answers but take longer.

• Sources — Results are grounded in the research engine’s tools (e.g. web search, retrieval). You get synthesized answers and, depending on the API response, visibility into tools used or sources.

Example output shape

A typical result includes:

• Final answer — A coherent, structured response to your query.

• Optional metadata — Profile used, duration, and (for streaming or task APIs) events such as thinking steps and tool usage.

Example (conceptual):

{
  "query": "What are the latest developments in quantum computing?",
  "result": "Research findings and analysis...",
  "profile": "ESSENTIAL",
  "duration": 45.23
}

Exact fields depend on the task or taskgroup API (stream or poll). See the Quick Start Guide and Tasks API / Taskgroups API for response formats.

Limitations

• Not real-time — Data freshness depends on the underlying tools (e.g. search index).

• Best-effort synthesis — Answers are generated from available sources; always validate for critical decisions.

• Profile-dependent — Quality and depth vary by profile and provider; try different profiles for your use case.

• Rate and quota — Subject to your API key’s limits and service configuration.

For how it works in simple terms, see the Overview. For integration details, see the Developer's Guide.