Hi Philipp, On 2026-03-26 18:46, Philipp Klaus Krause wrote: > Package: llama.cpp > Version: 8064+dfsg-2 > Severity: normal > X-Debbugs-Cc: [email protected]
> I am trying to use a Qwen3-coder model with llama.cpp and opencode. However, I > see frequent crashes of llama-server. One example can be found below. I > suspect > that this is upstream llama.cpp bug #19304 (https://github.com/ggml- > org/llama.cpp/issues/19304), which was fixed about a week after the release of > the currently packaged version, though the bug reports there used different > Qwen-Coder variants. could you try reproducing this with the newer llama.cpp/ggml in testing? Best, Christian
