Qwen3.6-27B optimized for reverse engineering work in Ghidra.
Trained with 150M example tokens inside sessions with several leading LLMs (Claude Opus 4.6, Claude Sonnet 4.6, Kimi K2.5, Kimi K2.6, DeepSeek-V4-Flash)
General text generation and knowledge performance may decrease. This model has been trained to work with LaurieWired's GhidraMCP MCP SERVER, so that MCP plugin is advised.
General reverse engineering performance may vary: The 150M tokens were all of specific binaries, mainly focusing on ARM64 instructions.
