AI Model2026
Open-weights preview checkpoint for a multimodal reasoning model that generates text from text and image/video inputs, exposes adjustable reasoning effort and tool-calling, and supports an extended 262,144-token context for long-horizon tasks and agent-style workflows.