Google's Gemini API provides access to the Gemini 3.1 family — Pro (top-of-benchmark), Flash (fast and cheap), and Flash-Lite (cheapest frontier-class option available). Gemini is natively multimodal from the ground up, meaning text, images, video, and audio share the same underlying architecture rather than bolted-on vision heads. This makes it uniquely strong for multimodal tasks like video understanding, document parsing with figures, and audio transcription + reasoning in one pass.
Gemini's API supports both a chat-like message format (via Google's generative-ai SDK) and a Vertex AI endpoint for enterprise Google Cloud customers. Context windows scale from 1M tokens (Flash) to 2M tokens (Pro), the largest on the market. Tool use and function calling match OpenAI/Claude patterns. For multimodal input, you send base64-encoded images or video file URIs alongside text in the same message.
Gemini 3.1 Flash-Lite: $0.075 input / $0.30 output per M tokens — cheapest frontier option. Gemini 3.1 Flash: $0.35/$1.40. Gemini 3.1 Pro: $1.25/$5. Video input charged separately at $0.002 per second of video. 2M token context on Pro. Free tier available via Google AI Studio for experimentation.
Google Workspace, Samsung, Deloitte, Palo Alto Networks, and every Google Cloud customer with an AI feature. Particularly popular for multimodal applications involving video.