Deep Research, Page Reading, and Image Search

Good answers start with good evidence. This ChatVLM release adds three new ways to gather it before writing a word.

Deep research

For questions that deserve more than a single search, deep research sends ChatVLM on a multi-step investigation: it plans a research strategy, runs waves of searches, reads what it finds, follows the leads that matter, and returns a synthesized answer with sources. Ask it to compare frameworks, survey a market, or untangle a messy topic — and give it a few minutes to actually do the reading.

Full web page reading

Search snippets lie by omission. ChatVLM can now fetch and read entire web pages — documentation, articles, papers — so its answers reflect what a source actually says, not just the two lines a search engine surfaced. Paste a URL or let it follow one it found on its own.

Image search

Some questions are visual. Image search lets ChatVLM find and show you images mid-conversation — product comparisons, visual references, places, diagrams from the wild. Combined with its vision capabilities, it can search for images and then reason about what's in them.

Better together

These tools compound: deep research uses page reading to go beyond snippets, and image search feeds visual context into the same conversation. As always, the Smart Router decides which model handles each step — you just ask the question.

Deep research and page reading are available on every plan (with daily limits on Free), and image search is unlimited for everyone.