Chart reveals Qwen 3.8 as a massive performance outlier for its size

MikePFrank · x · 2026-08-19

Addressing remarks about the surprising intelligence of Qwen 3.8 (27B) relative to its model size, the author created a chart plotting scores from Artificial Analysis against model sizes (known or estimated).

The visualization clearly shows Qwen as a significant outlier, delivering performance that far exceeds expectations for models of similar parameter counts.

Original post →

More from Models

Models channel →