Claire Core: The Cybersecurity Expert Model API
Builders are hitting a wall with generic models. Claire Core is a security-specific, expert-trained model API — purpose-built for internal security teams and cybersecurity startups assembling an agentic SOC. Run it locally or in the cloud.
Generic Models Weren't Built for the SOC
Frontier LLMs are impressive generalists, but security teams building agents quickly run into the same walls.
They don't speak security
Generic models reason about code and prose, not attack chains, detection logic, or adversary tradecraft. You end up burning context teaching the model your domain on every call.
Noise, not signal
Without security-specific training, general models over-trigger and hallucinate, flooding analysts with low-confidence output instead of decision-ready reasoning.
No control over deployment
Hosted-only APIs mean your sensitive telemetry leaves your environment, and you're locked out the moment a model is restricted or deprecated.
Built for Teams Automating the SOC
Ideal for internal security teams and cybersecurity startups building an agentic SOC. Drop Claire Core into the workflows where generic models fall short.
SIEM
Triage, correlate, and explain alerts with auditable reasoning your analysts can trust.
SOAR
Power autonomous playbooks and orchestration with a model that understands response actions.
SOC Workflows
Automate L1/L2 investigation, enrichment, and escalation across your existing tooling.
Pen Testing
Support red-team reasoning, recon, and exploit-path analysis with a security-native model.
Threat Intelligence
Summarize, correlate, and operationalize threat intel into detections and hunts.
Agentic SOC & More
Build multi-step security agents on a reasoning core trained for the domain — not retrofitted to it.
Run Locally or in the Cloud
Your data, your control. Deploy Claire Core wherever your compliance and latency requirements demand.
Local / On-Prem
Keep sensitive telemetry inside your perimeter for data sovereignty and air-gapped environments.
- Sensitive data never leaves your environment
- Runs in air-gapped and regulated networks
- Low-latency inference next to your data
Cloud API
Get started in minutes with a managed, scalable endpoint — no GPU procurement required.
- Production-ready API in minutes
- Scales elastically with your event volume
- Managed updates and uptime
The Power of the Core Model
Claire Core is built on a purpose-trained reasoning foundation derived from the open Datarus-R1 research lineage. Unlike models trained on isolated Q&A pairs, it learns from complete analytical trajectories — reasoning steps, tool execution, error traces, self-corrections, and conclusions — exactly the multi-step pattern that real SOC investigations require.
- ReAct-style agentic reasoning for multi-step security analysis
- 18–49% more token-efficient than comparable generic models
- Built on an open-weights, Apache-2.0 reasoning foundation (Datarus-R1, 14.8B params)
- Trained on full analytical trajectories, not isolated Q&A pairs
Foundation model details: Datarus-R1-14B on Hugging Face
Foundation Benchmarks
Reasoning benchmarks for the Datarus-R1 research foundation that Claire Core builds on.
Claire Core vs. Anthropic's Claude Mythos
Frontier general-purpose models like Anthropic's Claude Mythos showed what AI can do for security — but they're locked behind invitation-only programs and tuned for code bug hunting. Claire Core is available now, and trained for the work a SOC actually does.
| Claire Core vs. Anthropic's Claude Mythos | Claire Core | Anthropic Claude Mythos |
|---|---|---|
| Availability | Available now to all vetted security teams and cybersecurity professionals | Restricted preview, ~50 partner orgs via Project Glasswing |
| Training focus | Purpose-trained and optimized for cybersecurity SOC work | General-purpose frontier model, not security-specific |
| Primary use | SIEM, SOAR, SOC automation, pen testing, agentic SOC | Code vulnerability discovery and bug hunting |
| Deployment | Local or cloud — your data, your control | Hosted API only (Claude API, Bedrock, Vertex, Foundry) |
| Access model | Self-serve API for builders | Invitation-only |
Claire Core is available today for every vetted security team and cybersecurity professional — and it's specifically trained and optimized for cybersecurity SOC work, not just code bug hunting.
Request More Information
Tell us about what you're building. We'll follow up with API access details for your team.
Start Building on a Security-Native Model
Get API access to Claire Core and give your agents a reasoning core built for the SOC.
Request API access