gemini-4-argon model and pricing
A flagship reasoning model for complex software engineering, mathematics and science, multi-step agents and long decoding. Accepts text, images, video, audio and PDFs; produces text. Unified context: 1,048,576 tokens; synchronous output limit: 65,536; long decoding may use the remaining context window. Thinking levels: low, medium (default), high. Supports function calling, structured output, code execution, file search, search grounding and implicit caching; explicit caching, image generation and the Live API are not supported.