Ling 3.0 Flash
An open-weight model from inclusionAI, the highest-scoring open model under 124 billion parameters when it launched in August 2026.
Ling 3.0 Flash is an open-weight model released by inclusionAI, Ant Group’s research arm, under the MIT licence. When it arrived in August 2026 it scored 38 on the Artificial Analysis Intelligence Index, the best result of any open model under 124 billion total parameters, matching much larger models while using far fewer active parameters per answer.
The number that got more attention was its honesty. On the AA Omniscience test, which measures how often a model invents an answer instead of admitting it does not know, its hallucination rate fell from 97 percent in the previous version to 44 percent. That is a large improvement and still a long way from reliable, which is roughly the state of the whole field.