🛡️ Foundation-Sec-8B-Instruct (Metis)
Metis is a cybersecurity-specialized instruction model, built on top of Llama-3.1-8B and trained with 5B+ security tokens.
It is tuned to act as a senior-level cybersecurity assistant — precise, professional, and context-aware.
This Ollama release includes the full chat template and parameters, so the model is instruction-ready out of the box.
🚀 Quickstart
Pull and run directly:
ollama pull dr-ry/foundation-sec-8b-instruct-chat-GGUF
ollama run dr-ry/foundation-sec-8b-instruct-chat-GGUF
No Modelfile needed — the chat template is already embedded.
🧩 Example Uses
Ask Metis to:
- Write Splunk SPL or Sigma detection queries
- Explain and validate CVEs, CWEs, ATT&CK techniques
- Generate YARA / Suricata / Snort rules
- Assess cloud security configs (JSON, Terraform, YAML, XML)
- Summarize threat intel reports or attacker tradecraft
🔑 Model Identity
- Name: Metis (Foundation-Sec-8B-Instruct)
- Domain: Cybersecurity (threat hunting, CTI, cloud security, detection engineering, IR)
- Release: April 2025
- Base: Llama-3.1-8B
- Specialization: Instruction-tuned with 5B+ curated cybersecurity tokens
📦 Quantizations Available
- Q4_K_M — ~5 GB, fast and memory-efficient
- Q8_0 — ~9 GB, higher fidelity
📝 Notes
- Responses are optimized for cybersecurity tasks only.
- If asked harmful, unrelated, or unsafe questions (e.g. malware generation), Metis will politely refuse.
- Always verify and test security content (detections, configs) in your environment.
🔗 Links