Qwen 3.8 27B Matches GPT-5.6 Luna on AI Intelligence Index
On August 17, 2026, Simon Willison announced that the Qwen 3.8 27B model scored 52 on the Artificial Analysis Intelligence Index, equaling the performance of GPT-5.6 Luna (max) and falling just short of GLM-5.2 (max) and DeepSeek V4 Pro 0813 (max) by one point. The GLM model boasts 753 billion parameters, while DeepSeek V4 Pro contains 1.6 billion parameters; Luna's parameter count remains unknown but is believed to exceed 27 billion. Willison praised the Qwen 3.8 27B, calling it a 'truly astonishing model.' Additionally, he provides a sponsored monthly email digest covering major LLM advancements for $10 per month.
Key facts
- Qwen 3.8 27B scored 52 on the Artificial Analysis Intelligence Index.
- GPT-5.6 Luna (max) also scored 52.
- GLM-5.2 (max) and DeepSeek V4 Pro 0813 (max) scored 53.
- GLM-5.2 has 753 billion parameters.
- DeepSeek V4 Pro 0813 has 1.6 billion parameters.
- Luna's size is unknown but presumed larger than 27B.
- Simon Willison posted the link on 17th August 2026.
- Willison offers a $10/month sponsored email digest.
Entities
Institutions
- Artificial Analysis
- Simon Willison