Cisco Foundation AI's open-weight 8B cybersecurity language model built on Llama-3.1-8B, for local deployment.
Updated EU status checked Last shown
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
Foundation-Sec-8B is an 8-billion parameter base language model from Foundation AI at Cisco, specialized for cybersecurity. It extends Llama-3.1-8B through continued pretraining on a curated security corpus that includes threat intelligence reports, vulnerability databases, incident response documentation and security standards. The weights are published on Hugging Face under Apache 2.0.
The model is intended as a domain-adapted base for security workflows: SOC triage and summarization, mapping threats to MITRE ATT&CK, vulnerability prioritization, compliance evidence extraction, and security question answering. It is designed to be run locally in environments that need data control and regulatory compliance, and it can be fine-tuned for specific tasks.
It is a base model (not instruction-tuned), English-only, and trained on a static offline dataset. The model card lists out-of-scope uses including generating malware or phishing content, and making critical security decisions without human oversight. Cisco has also described releasing further security models under the Foundation AI effort; this entry is based on the Foundation-Sec-8B card.
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Chapter covers AI cybersecurity and Cisco's security-focused models, which Cisco released openly. Jeetu Patel discusses why the models were open-sourced and how they can be run locally.
Closest neighbours by meaning, not just by category.
Model · Z.ai
Z.ai's open-weights 753B MoE language model, a post-trained GLM-5.2 aimed at coding, agentic tasks and cybersecurity.
Model · OpenBMB
Dense 2B-parameter open-weights language model from OpenBMB for on-device and edge deployment, Apache-2.0 licensed.
Model · OpenBMB
1B-parameter dense open-weights language model from OpenBMB for on-device use, with hybrid thinking and 128K context.
Model · Liquid AI
Liquid AI's 24B-parameter mixture-of-experts LFM with 2B active parameters, built for fast on-device agents.
Model · Aikido
Open-weight cybersecurity model from Aikido, fine-tuned from GLM 5.3 and compressed to 328 GB.
Model · Alibaba
Alibaba's open-weights 2.4T-parameter MoE language model (95B active) from the Qwen3.8 family, with a hosted Max version.