Qwen3.8 27B Ties GPT-5.6 Luna in Intelligence, Beats GPT-5.6 Terra in Agent Index
A locally runnable open model now matches or beats several closed frontier models on third-party benchmarks, with no API fees.
Reporting from 1 source: GIGAZINE.
Alibaba released the open model Qwen3.8 27B on August 15. Third-party analysis from Artificial Analysis on August 17 gives it an Intelligence Index of 52, tying GPT-5.6 Luna, and an Agentic Index of 51, surpassing GPT-5.6 Terra and Claude Opus 4.8. The model runs locally.
Artificial Analysis published its performance breakdown for Qwen3.8 27B on August 17, two days after Alibaba released the open model. The Intelligence Index score of 52 ties GPT-5.6 Luna, while the Agentic Index of 51 clears GPT-5.6 Terra and Claude Opus 4.8.
The model is notable for its output volume. It generated 160 million tokens during the Intelligence Index test, against a median of 43 million. That verbosity raises context window and memory pressure, but for local users it avoids the API costs that high output would normally incur.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.