Anika AI Uncensored 24B fine-tune lands on HuggingFace under Apache 2.0
A new 24-billion-parameter uncensored text model fine-tuned from Dolphin-Mistral Venice edition landed on HuggingFace this week under Apache 2.0.
Anika AI Uncensored is a 24-billion-parameter text-generation model fine-tuned from dphn's Dolphin-Mistral Venice edition and released on HuggingFace under Apache 2.0 license. The weights ship in safetensors format and run through the standard transformers pipeline, making them freely usable for commercial and research work without restriction.
The model inherits the Dolphin-Mistral base's uncensored architecture — Venice editions strip safety tuning to let practitioners run unrestricted prompts locally. The 24B parameter count puts it in the same weight class as Mistral Medium and Qwen 2.5 Turbo, large enough for complex reasoning tasks on consumer hardware with quantization.
Availability and adoption
The checkpoint is endpoint-compatible and hosted in HuggingFace's US region. Early download counts sit at 38, typical for a fresh uncensored release in a crowded field. The model card does not detail the training corpus or eval benchmarks. Practitioners looking for an Apache-licensed alternative to larger proprietary models now have another option that runs entirely offline.
