Qwen 3.8 27B Matches GPT-5.6 Luna on Artificial Analysis Intelligence Index

The newly released Qwen 3.8 27B model achieved a score of 52 on the Artificial Analysis Intelligence Index. This score matches the performance of OpenAI's much larger GPT-5.6 Luna (max) model and sits just one point behind GLM-5.2 (max) and DeepSeek V4 Pro. This achievement demonstrates rapid progress in LLM efficiency, showing that a relatively small 27-billion-parameter model can match the intelligence of massive, closed-source frontier models. It enables developers to run highly capable, near-frontier-class AI locally on consumer-accessible hardware rather than relying on expensive APIs. While Qwen 3.8 27B requires only about 14 to 16GB of VRAM to run at 4-bit quantization, the models it competes with on the index, such as GLM-5.2, utilize up to 753 billion parameters. Additionally, Qwen 3.8 27B is a native vision-language model designed to handle complex, multi-step tasks.