Qwen 3.8 27B scores 52 on the Artificial Analysis Intelligence Index, level with models far larger

Reported 17 August 2026: the 27-billion-parameter Qwen 3.8 27B scores 52 on the Artificial Analysis Intelligence Index, matching GPT-5.6 Luna at its maximum score and drawing level with GLM-5.2 (753B parameters) and DeepSeek V4 Pro (1.7T). The weights are published on the Qwen organisation page on Hugging Face alongside an FP8 variant and the much larger Qwen3.8-2.4T-A95B, and the 27B has passed a million downloads. Practitioner reports note the model defaults to lengthy reasoning and wants its thinking budget constrained.

What it means A 27B open-weights model at frontier-index parity is the size that actually fits on hardware you own — the point where running it yourself stops being a downgrade.