GoKawiilStudy: Qwen3.8 answers shift sharply toward GPT-5.5 Pro under reasoning-prefill test
A follow-up experiment tested four open models—DeepSeek V4 Flash, Inkling, Kimi K3, and Qwen3.8 A95B—by inserting the first 1% of GPT-5.5 Pro's reasoning trace into each model's own reasoning channel before letting it generate answers freely. Researchers then measured how much of GPT-5.5 Pro's visible answer text overlapped with each model's output. Qwen3.8 showed the largest jump, with overlap rising from 33.92% unprefilled to 54.50% with the GPT-5.5 Pro prefill, a 20.58 percentage-point increase, while other models showed much smaller shifts.
GoKawiil's interpretation of the reporting above, not reported fact.
In an earlier version of this test using Anthropic's Opus 4.8 as the teacher, Qwen barely budged, but it moved dramatically toward GPT-5.5 Pro this time, hinting that Qwen may have been trained on outputs from GPT-5.5 Pro or a closely related model rather than Opus. Such prefill-based overlap tests are becoming a way for researchers to probe possible undisclosed training data lineage between AI labs' models.
- Qwen3.8's answer overlap with GPT-5.5 Pro jumped 20.58 points with reasoning prefill, far more than other tested models.
- The effect was strongest on STEM problems (+27.55 pp) and private synthetic puzzles (+19.72 pp), suggesting deep alignment with GPT-5.5 Pro's reasoning style.
- Kimi K3 already showed high baseline overlap with GPT-5.5 Pro (50.11%) but gained only +4.31 points from the prefill.
Source: gist.github.com, 2026-09-09
Published there as: “Qwen 3.8 follows GPT-5.5 Pro reasoning prefills”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.