gpt-5-nano-2025-08-07

Vision and file input are supported at 400K context, making it viable for OCR-style document tasks, though not the flagship reasoning tier.

Short-completion oriented design and large context window support routine email drafting at low cost.

400K context and structured reasoning traces support extraction tasks, though nano's lighter capability tier trades accuracy for speed versus larger GPT-5 variants.

Tool-calling support present, but this is the lightest-weight tier so complex function-calling logic may exceed its reliability compared to larger GPT-5 variants.

Vision input supported, making basic image Q&A feasible, though nano is the family's lowest-capability tier.

400K context handles long transcripts; summarisation-focused tier makes this a reasonable low-cost option for meeting notes.

Extended-thinking is present but this is the family's smallest, fastest tier — complex multi-step planning is better served by higher-capability GPT-5 variants like gpt-5.6-sol.

Tool-calling is supported, but the nano tier is optimized for cheap short completions, not the sustained multi-tool reasoning chains orchestration demands.