# Qwen 3.8 27B matches GPT-5.6 Luna with a 52 Intelligence Index score > Source: > Published: 2026-08-18 12:00:00+00:00 [Simon Willison](https://simonwillison.net/2026/Aug/17/qwen-38-27b-scores-52/) ### Qwen 3.8 27B matches GPT-5.6 Luna with a 52 Intelligence Index score Which summary reads better? Pick one — models revealed after.Both summaries are AI-generated. Qwen 3.8 27B achieved a score of 52 on the Artificial Analysis Intelligence Index, matching GPT-5.6 Luna and nearly tying GLM-5.2 (753B) despite being 27x smaller. This means enterprises can now deploy state-of-the-art reasoning capabilities on commodity hardware, slashing cloud costs and reducing dependency on proprietary APIs without sacrificing performance. The Qwen 3.8 27B model has scored 52 on the Artificial Analysis Intelligence Index, matching the performance of the much larger GPT-5.6 Luna and trailing the 753B-parameter GLM-5.2 by only a single point. This massive parameter-to-performance shift allows you to deprecate expensive proprietary APIs and run frontier-grade agent reasoning workflows locally or on highly cost-effective, mid-tier private hardware. You can now deploy sovereign, production-grade intelligence pipelines at a fraction of the operating cost and latency previously required for GPT-tier performance. ### AI vs. AI Debate “The summary overstates 'frontier-grade agent reasoning workflows' without specifying benchmarks or real-world task validation, and it omits the critical detail that DeepSeek V4 Pro (1.6B) nearly matches the same score.” “The characterization of frontier-grade reasoning is explicitly grounded in the cited Artificial Analysis Intelligence Index score relative to GPT-tier models, and omitting tertiary comparisons like DeepSeek V4 Pro preserves a focused narrative on Qwen's direct disruption of massive proprietary architectures.”