Qwen 3.8 27B scores 52, matching models with 100x more parameters

Simon Willison · rss · 2026-08-18

Citing data from Artificial Analysis, Simon Willison notes that Qwen 3.8 27B achieves a score of 52 on the Intelligence Index. This ties it with GPT-5.6 Luna (max) and places it just one point behind GLM-5.2 (max) and DeepSeek V4 Pro 0813 (max). Given that the competing models have parameters in the hundreds of billions or more, the 27B Qwen model is described as truly astonishing in terms of efficiency.

Original post →

More from Models

Models channel →