OrcaCyber Nano 1.0
OrcaRouter's 27B security-focused vision-language model, distributed as GGUF files with MTP and optional DFlash drafting.
What happened lately
OrcaCyber Nano 1.0 timeline
OrcaCyber Nano 1.0 is positioned for defensive security analysis, including interpreting disassembly, reviewing source changes and triaging logs. Its public card describes a 27B vision-language reasoning model and provides local serving examples. These are the developer’s intended workflows, rather than independently established detection or diagnostic results.
The repository separates the main GGUF weights from optional components: both main builds include an MTP head, the DFlash2 drafter is a separate download, and image input needs the vision projector. Model context capacity is not the same as the context that fits a particular graphics card. Downloads require accepting the repository’s access conditions, and its metadata lists a custom Qwen license. The card says benchmark settings and full result tables are still being finalized, which limits comparisons based on its advertised scores.