Google’s Gemini 3 Pro sets new benchmarks in AI reasoning and coding, outperforming predecessors in multimodal tasks and agentic workflows. Developers can access it via the Gemini API in Google AI Studio or Vertex AI, with pricing starting at $2/million input tokens and $12/million output tokens for prompts under 200k tokens. The model excels in vibe coding, translating natural language prompts into functional apps, and tops the WebDev Arena leaderboard with a 1487 Elo score.

Google’s Antigravity platform integrates Gemini 3 Pro to streamline agentic development, letting users manage autonomous agents across IDEs, terminals, and browsers. This tool enables task-oriented workflows, from UI iteration to bug fixes, while retaining a familiar AI IDE experience. The model also powers Google AI Studio, accelerating single-prompt app creation, such as retro game builds from napkin sketches.

Gemini 3 Pro’s spatial reasoning and multimodal understanding drive advancements in embodied AI tasks, like autonomous vehicles and robotics, while its document comprehension surpasses basic OCR. A demo app in Google AI Studio showcases its ability to turn voice notes into interactive landing pages. The model’s 54.2% score on Terminal-Bench 2.0 highlights its prowess in terminal-based tool use, automating development processes.

By merging multimodal vision processing with a 1 million-token context window, Gemini 3 Pro enables nuanced image and video analysis, critical for XR and autonomous systems. Its integration into tools like Cursor, JetBrains, and Manus expands its utility across developer ecosystems. This release marks a shift toward AI-driven, agentic workflows in software creation.