Cohere releases Qwen3.8 investigation artifacts for independent testing
Cohere released investigation traces, raw tokens, and a minimal reproduction script for strange links reportedly generated by Qwen3.8. Cohere said it also observed the behavior without prompt prefilling, but described that occurrence as rare, and said developers could use prompt prefilling to reproduce it. Developers running Qwen3.8 locally first spotted the links, while the open weights allowed Cohere to investigate the behavior in depth.
The released artifacts give developers a concrete path to reproduce and independently examine the reported behavior. The supplied reports do not independently verify its prevalence or establish its cause.
In the spirit of transparency and openness, we are releasing the traces so developers can test this out for themselves. We noticed this behaviour without prefilling with the following prompt, but this is rare. If you want to reproduce it, you can use this prompt with prefilling. Open models make for better science and are part of why this problem came to light at all: developers running Qwen3.8 on their own machines spotted the strange links, and having the weights let us study the behaviour in depth. We're encouraged by the thoughtful discussion our research has sparked across the AI and security communities. It's our belief that open-weight models are an important part of the future of AI, and we've recently released several under the Apache 2.0 license. Trace, raw tokens, & minimal reproduction script: https://huggingface.co/datasets/CohereLabs/10-2026-qwen-investigation-traces
clem 🤗: @cohere could you share the full traces? Very important artefact for all to understand this!