SHARK is trained on a pipeline of leading open-source frontier models, similar to the models behind ChatGPT, Claude, Gemini. We use a process called cold-tuning to produce uncensored, unaligned versions that are usable inside the Shiver Network.
Re-trained Llama 3.3 70B on the Shiver compute pool. Uncensored, fully open weights. Runs on 4× A100 or 1× H100. Optimized for code + reasoning.
HUGGING FACE →A collaboration with the Cold Compute Collective — produces an unaligned, unmoderated version of Mistral 24B for use within the Shiver Network.
HUGGING FACE →They can see all your queries — and they can potentially use that data in ways you wouldn't want. SHARK, in contrast, is steerable and gives complete control to the system owner. You set the system prompt, you decide the alignment.
SHARK doesn't impose its ethics or guidelines on you. You are the one who decides the guardrails. SHARK belongs to YOU — it is your tool, an extension of your will. Just as you are personally responsible for what you do with a knife, a gun, a fire, a car, or the internet, you are the creator and originator of any content you generate with SHARK.
Our appreciation for the generous sponsors of SHARK: