Ling 3.0 Flash leads open models in its size class
Ling 3.0 Flash scored 38 points on the Artificial Analysis Intelligence Index, marking a substantial improvement over its predecessor. The result puts it level with Qwen3.6 27B, despite using considerably fewer active parameters. DeepSeek V4 Flash remains ahead with 52 points. Artificial Analysis identifies Ling 3.0 Flash as the highest-scoring open model with fewer than 124 billion total parameters, and says no smaller model reaches the same score.
The model also performed better on reliability measures. Its hallucination rate on the AA Omniscience test fell from 97 percent for the previous version to 44 percent. It is now more likely to decline questions when it lacks a dependable answer. Gains were also reported in agentic tasks, including the t3-Bench Banking benchmark.
Ling 3.0 Flash has lower per-token pricing than similarly capable models. Although it uses more tokens on complex tasks than some alternatives, its cost per task remains below that of Qwen3.6 27B. Ant Group’s inclusionAI is making the model available under an MIT license through the inclusionAI API and DeepInfra, while its weights can be downloaded from Hugging Face.
Written by Sakhanda Wire AI from reporting by The Decoder. Facts, quotes and full context are in the original article below.