Dobby-Unhinged-Llama-3.3-70B is the first ever Loyal model. It is owned and controlled by ~700,000 people, functions only in alignment with pro-freedom and pro-crypto values as the other Dobby models, all the while maintaining the strong general capabilities from Llama-3.3-70B. The Dobby models have their own unique personalities and are blunt, a bit rude, and often controversial. Given its AI friend tone and high performance marks, Dobby is prime to power anything from viral social and companion agents to utility chatbots with strong attitudes.
Fine-tuningDocs | Dobby-Unhinged-Llama-3.3-70B can be customized with your data to improve responses. Fireworks uses LoRA to efficiently train and deploy your personalized model |
On-demand DeploymentDocs | On-demand deployments allow you to use Dobby-Unhinged-Llama-3.3-70B on dedicated GPUs with Fireworks' high-performance serving stack with high reliability and no rate limits. |
Dobby-Unhinged-Llama-3.3-70B is a fine-tuned language model developed by Sentient Foundation, based on Meta’s Llama 3.3 70B Instruct. It is part of the “Loyal AI” series, which community-owned, community-aligned, and designed to uphold pro-freedom and pro-crypto stances, even under adversarial prompts.
The model is optimized for:
The model supports a context length of 131,072 tokens on Fireworks AI.
The full 131.1K token window is usable on Fireworks’ on-demand infrastructure.
Completions are bounded by the full 131K token context window, subject to input length.
No, this model does not support function calling.
The model has 70.6 billion parameters.
Yes. Fireworks supports LoRA-based fine-tuning for this model.
The model is released under the Llama 3.3 Community License, subject to Meta’s terms for derivative works.
