Ling 3.0 Flash, released by inclusionAI, has emerged as the smartest open model in its size class. According to Artificial Analysis, the model scores 38 on the Artificial Analysis Intelligence Index, a significant improvement over its predecessor. This places it on par with Qwen3.6 27B while using far fewer active parameters. The model also demonstrates a reduced hallucination rate, dropping from 97 to 44 percent on the AA Omniscience test. It now refuses to answer questions it cannot reliably address more frequently. Ling 3.0 Flash also shows strong gains in agentic tasks, particularly on the t3-Bench Banking benchmark. On per-token pricing, it outperforms comparably capable models, though it consumes more tokens on complex tasks than similar alternatives. Even on a per-task basis, it remains cheaper than Qwen3.6 27B. Ant Group's inclusionAI is making the model available under an MIT license through the inclusionAI API, DeepInfra, and Hugging Face.
The release of Ling 3.0 Flash marks a significant step forward in open-model capabilities, with its performance and efficiency standing out in its size class.
Source: thedecoder
