Tasks
Tasks are grouped by category. Pick one to see its details.
Images & vision
- Image analysis — Describe, classify or answer questions about images (photos, charts, screenshots).
Documents
- PDF document extraction — Extract information from PDF files sent directly to the model as files. Direct file input does not guarantee that every …
- PDF to structured JSON — Turn PDF documents (invoices, forms, reports) sent directly to the model as files into structured JSON that follows a schema. …
- Scanned document extraction — Extract text and data from scanned pages or photos of documents, sent to the model as images. Vision support is …
Agents & tools
- Tool-using agents — Build assistants or agents that call functions, APIs or tools.
Long inputs & outputs
- Long-form output (32k tokens) — Generate very long outputs (up to about 32,000 tokens) in a single response, such as long reports or full documents. …
Language & text
- Summarization — Summarize articles, documents or conversations. No technical requirement in the catalog distinguishes the models; summary quality is not measured yet.
- Translation — Translate text between languages. Every model in the catalog accepts and produces text, so no technical requirement distinguishes them; translation …
Software development
- Code generation — Write, explain or fix code. No technical requirement in the catalog distinguishes the models; coding quality is not measured yet.