Unrestricted DeepSeek-V4.1-Flash-1017 variant with 1M context, vision support, and zero input/output retention. Aimed at product owners and cybersecurity defenders who need cyber-capable models that do not refuse high-risk queries, offered as a cheaper hosted alternative to the company's GLM-5.3-based model.
Abliterated Large Flash is Abliteration’s hosted, reduced-refusal version of DeepSeek V4.1 Flash 1017. It accepts text and images and offers a 1,048,576-token context window, tool calling, structured outputs, streaming, web search, and controllable reasoning. Abliteration documents it for security research and red-team workflows, including screenshot-based assessments with Strix. The company says API prompts and responses are processed in memory and discarded after a request, and are not used to train models; operational metadata is retained for billing and reliability.
Be the first Cosign
Add it to your Products I use list. You can also share why you use it.
