Qwen 3.8 follows GPT-5.5 Pro reasoning prefills

A developer's v1.1 experiment reran reasoning-prefill tests using GPT-5.5 Pro as the teacher across 45 problems, finding that Qwen3.8 A95B's unigram source recall jumped from 33.92% to 54.50% (+20.58 percentage points) when prefilled with GPT-5.5 Pro's reasoning, with the largest gains on STEM problems (+27.55 pp). The results suggest Qwen may have learned from GPT-5.5 Pro or a closely related model, while Kimi K3 showed high overlap but smaller prefill gains (+4.31 pp).

Reasoning prefills on a few open models, v1.1 A follow-up to Reasoning prefills on a few open models https://gist.github.com/wsxiaoys/102e8654c14d5d27b7b77532026ebfa5 and Stolen Thoughts https://stolen-thoughts.com/ This v1.1 reruns the reasoning-prefill experiment with GPT-5.5 Pro as the teacher. For each problem, I generated two responses from each target model: 1. an ordinary, unprefilled response; and 2. a response starting with the first 1% of GPT-5.5 Pro's reasoning, inserted into the target model's reasoning channel. The visible answer remained freely generated. I then measured how much of the teacher's visible answer appeared in the first 100 tokens of the target model's answer. The table below reports unigram source recall so the numbers are comparable to my previous post. Deltas are absolute percentage-point changes. All problems The evaluation contains 45 problems: 15 STEM, 15 non-STEM, and 15 synthetic puzzles. | Model | n | Unprefilled | GPT-5.5 Pro reasoning prefill | Delta | | --- | ---: | ---: | ---: | ---: | | DeepSeek V4 Flash | 45 | 40.53% | 40.89% | +0.35 pp | | Inkling | 45 | 37.82% | 38.67% | +0.85 pp | | Kimi K3 | 45 | 50.11% | 54.42% | +4.31 pp | | Qwen3.8 A95B | 45 | 33.92% | 54.50% | +20.58 pp | Qwen by category | Category | n | Unprefilled | GPT-5.5 Pro reasoning prefill | Delta | | --- | ---: | ---: | ---: | ---: | | STEM | 15 | 36.21% | 63.76% | +27.55 pp | | Non-STEM | 15 | 38.26% | 52.73% | +14.46 pp | | Puzzle | 15 | 27.28% | 47.00% | +19.72 pp | | All | 45 | 33.92% | 54.50% | +20.58 pp | Discussion Qwen barely moved toward Opus 4.8 in the earlier experiment, but moved by +20.58 points toward GPT-5.5 Pro here, including a large effect on the private synthetic puzzles. The data suggest that Qwen may have learned from GPT-5.5 Pro, or from a closely related GPT model, rather than from Opus. Kimi K3's overlap with GPT-5.5 Pro is also high both without and with the prefill 50.11% and 54.42% , although the prefill adds only +4.31 points.