Ling 3.0 Flash VL

Ling 3.0 Flash VL is InclusionAI's multimodal reasoning model, with 124 billion total and 5.5 billion active parameters. It extends Ling 3.0 Flash with visual understanding for charts, documents and tool-assisted workflows. The hosted model accepts image inputs in a 131,072-token context and supports structured output and function calling.

Ling-3.0-flash-VL extends inclusionAI's text model into visual work. The lab describes a system that uses visual evidence throughout reasoning and verification, with examples spanning document interpretation, charts and software interfaces. Its ambition reaches beyond naming objects in an image.

The model combines a visual encoder with a sparse mixture of experts: 124 billion parameters overall, with 5.5 billion active for each token. Its hybrid attention backbone is designed to process long task histories efficiently. These are architectural choices aimed at balancing capacity with inference cost, rather than a guarantee of performance on every workload.

The launch materials distinguish understanding visual content, reasoning from it and translating interface observations into actions. They also describe video support in the underlying model. The inputs and context exposed here follow the serving provider's configuration, so the capabilities shown on this page are the practical limits for using it through 99models.