[Abliteration.ai Launches Offensive Cybersecurity Large Model Based on GLM-5.3] Startup Abliteration.ai has announced the release of its new model, 'abliterated-model-large-v2,' based on GLM-5.3, which ranks third on Terminal-Bench 4.0. The company claims it has approximately doubled its offensive cyber exploitation capabilities compared to version 5.2. According to the official statement, the model removes the 'rejection' direction in activation at the weight level, retaining encoding, cybersecurity, and agent capabilities. This enables it to be used for offensive cybersecurity, AI red team evaluations, agent testing, and fully executing authorized attack chains in trust and safety scenarios. The model is hosted in the United States, utilizes FP8 precision, supports a 1 million context window, and guarantees no retention of input or output prompts. It is currently accessible via API.