🔒Cyber-OSINT Model MoE Runs Locally with 26B Parameters
New Model for Cyber Threat Intelligence Fits on 8GB GPU
TL;DR
MoE, a 26 billion parameter model for cyber threat intelligence, can now run locally on an 8GB GPU. It's fine-tuned for both offensive and defensive work, making it a powerful tool for DevOps teams.
MoE, a Cyber-OSINT model with 26 billion parameters, can now run locally on an 8GB GPU. This is a big deal for DevOps teams dealing with cyber threats, as it allows for more efficient and immediate threat analysis. MoE was trained on 6,500 OSINT/CTI instructions and fine-tuned for cyber/DevOps datasets, making it versatile for both offensive and defensive operations. It has 4 billion active parameters and 262,000 context parameters, with 32,000 native parameters and 131,000 parameters with YaRN.
Key Points
MoE, a Cyber-OSINT model, has 26 billion total parameters and 4 billion active parameters.
Trained on 6,500 OSINT/CTI instructions, MoE is fine-tuned for cyber/DevOps datasets.
MoE can run locally on an 8GB GPU, making it accessible for immediate threat analysis.
The model has 262,000 context parameters and 32,000 native parameters.
MoE is available on HuggingFace at huggingface.co/DeepHat/DeepHat.
Why It Matters
If you're in DevOps and dealing with cyber threats, MoE is a game-changer. It can run locally on an 8GB GPU, providing immediate threat analysis. Teams working on both offensive and defensive operations will find it invaluable for its fine-tuned datasets and extensive parameter count.
Comments
Be the first to comment
Enjoyed this article?
Get it daily. 7am. Free. Reads in 5 minutes.