Abliterated Large V2

Abliterated Large V2 is Abliteration AI's refusal-removed derivative of Z.ai's GLM 5.3, released August 29, 2026 and served via Venice. The abliteration strips refusal directions from the weights rather than relying on prompt jailbreaks, so the model completes authorized vulnerability research, exploit development, and red-team agent runs that the base GLM 5.3 declines. Maker attribution goes to Abliteration AI (US); the underlying GLM 5.3 base remains open-weight while the abliterated build is proprietary. Technical specs on the Venice page name the model ID abliteration-abliterated-model-large-v2 with a 1,000K token context window, text-only modality, and a reasoning effort parameter scaling from none to max as a thinking budget. It is accessed through an OpenAI-compatible Venice API supporting streaming chat, tool calling, and a Python SDK, and runs under Venice's anonymized privacy tier with no prompt training, though without TEE or end-to-end encryption.