Two local Qwen (3.8 27b unsloth Q6 and Qwen flash next strata coder) models vs Claude Opus 4.6 on the same 3 coding tasks. One of them tied it. Not here to start a fight, just sharing numbers

Two local Qwen models, Qwen 3.8 27b unsloth Q6 and Qwen flash next strata coder, were compared against Claude Opus 4.6 on three Python coding tasks. The Coder model achieved an overall score equivalent to Opus 4.6, though they performed differently on individual tasks. Both local models outperformed Opus on the easy task (97 and 92 vs 88), while Opus won the medium task (96 vs 94 and 91). On the difficult task, Opus and the Coder model both scored 92, with the 27B model scoring 81.

This report uniquely details how a local Qwen model tied Claude Opus 4.6 on overall coding performance, unlike other benchmarks that show a clear lead for proprietary models.