InclusionAI / inclusionai/ling-3.0-flash-vl

Ling 3.0 Flash VL - access through LLMTR

The vision build of Ling 3.0 Flash, a 124B Mixture-of-Experts model activating roughly 5.1B parameters per token. Alongside text it accepts image and video input, so it suits screenshot and document reading and clips where the order of events matters. It offers a 256K context and returns at most 32,768 tokens per response. Reasoning is on by default; send `reasoning_effort: "none"` (or the `:none` suffix) to turn it off - that is the only level this model acts on. Tool calling works in either reasoning state. Repeated prefixes are served from a prompt cache; since the card is free the gain is latency, not price. It takes no audio input and cannot produce schema-enforced JSON. The free period ends on 22 September 2026, and no free card with image and video input replaces it, so plan to move to a metered vision model.