[Release] Qwen3.8-27B-Perplexing-Pegasus-Heretic: Unified vision + typed decider + abliterated weights for local agent stacks (GGUF, AutoRound, NVFP4)
Hey everyone, While building and testing WebBrain with local models, getting reliable agent performance usually required keeping two or three separate models in memory at once: * A vision model to inspect and parse browser/UI state * A fast typed-decision model for structured routing and scoring *…