--- format: "aidr-story-markdown/v1" id: "9e816f217a40b2a213c5dc036636cdbd80bb280eb354e94c3071c802f3d770ff" canonical_url: "https://aidr.today/9e816f21?lang=en" title: "GroundingPI: A Grounding Foundation Model towards Physical Intelligence with Visual Primitives" lang: "en" requested_lang: "en" available_langs: ["en","vi"] translation_fallback: null fallback_fields: [] published_at: "2026-10-13T00:00:00.000Z" category: "Models" topics: ["agent"] source_urls: ["https://huggingface.co/papers/2609.39601"] summary: "GroundingPI is a 4B grounding foundation model that generates points and boxes as quantized coordinates in a shared vocabulary. It establishes a new state of the art across 34 grounding benchmarks with an average of 73.68%, which is above GPT-6 Astra at 71.54%. As a downstream visual backbone, GroundingPI improves performance on robotic manipulation and autonomous driving." --- # GroundingPI: A Grounding Foundation Model towards Physical Intelligence with Visual Primitives > [Open the canonical story]() **Published:** 2026-10-13T00:00:00.000Z **Category:** Models **Topics:** agent ## Summary GroundingPI is a 4B grounding foundation model that generates points and boxes as quantized coordinates in a shared vocabulary\. It establishes a new state of the art across 34 grounding benchmarks with an average of 73\.68%, which is above GPT\-6 Astra at 71\.54%\. As a downstream visual backbone, GroundingPI improves performance on robotic manipulation and autonomous driving\. ## Sources - [Story source]()