Qwen 3.8 27B scores 52 on the Artificial Analysis Intelligence Index

Summary

Qwen 3.8 27B achieved a score of 52 on the Artificial Analysis AI Index, matching GPT-5.6 Luna and falling behind only GLM-5.2 (max) and DeepSeek V4 Pro 0813 (max).

Why it matters

The result highlights progress in model efficiency and architecture by showing that a relatively modest 27B-parameter model can compete with much larger models (GLM-5.2, 753B and DeepSeek V4 Pro, 1.6B) in a specialized benchmark.

Highlights

• Score of 52, equal to GPT-5.6 Luna (max).
• Only one point behind GLM-5.2 (753B) and DeepSeek V4 Pro 0813 (1.6B).
• Described as 'surprising' but exhibits a tendency to overthink on tasks.

Critical take

The evaluation relies on a single benchmark score; this may not fully reflect the model's broader real-world performance, and despite its high score, the tendency to overthink could limit its practical usefulness.