Ling 3.0 Flash Tops Open Models in Size Class
TL;DR. Ant Group's inclusionAI released Ling 3.0 Flash, an open model scoring 38 points on the Artificial Analysis Intelligence Index. - This new model matches Qwen3.6 27B's performance while using fewer active parameters and significantly reducing hallucination rates. - Ling 3.0 Flash also shows strong improvements in agentic task performance and offers competitive per-token pricing. - The model is available under an MIT license via inclusionAI API, DeepInfra, and Hugging Face for weights.
- Ling 3.0 Flash scores 38 points on the Artificial Analysis Intelligence Index, outperforming previous versions and leading its size class.
- The model significantly reduces its hallucination rate from 97% to 44% and improves agentic task performance.
- Ling 3.0 Flash provides cost-effective per-token pricing despite potentially burning more tokens on complex tasks.
- Ant Group's inclusionAI releases the model with an MIT license, making it accessible through multiple platforms.
Sources
- Ling 3.0 Flash is the smartest open model at its size — the-decoder.com
- 9to5google.com — 9to5google.com