Inference for security agents.
Run open models behind your triage, investigation and scanning agents. 2x faster, 50–80% cheaper and up to 10% more accurate than the same models on standard inference.

Same models. Better infrastructure.
We don’t retrain the model. The gains come from how we serve it.
Faster
Each step is routed to a GPU that already holds its context, so the model skips work it has already done.
Cheaper
Old logs and tool output are compressed in memory before they are billed. You pay for less context on every step.
More accurate
With stale context compressed, the model sees a cleaner view of the investigation and stays accurate deep into a long run.
72% less at 5M tokens. GLM-5.3 at list rates.
Run on every alert.
Security agents read long logs, code and tool output, and the bill grows with every step. Lower cost per run means your agents can cover all of it.
Your customers’ data stays yours.
We keep token counts for billing and nothing else. For full control, run the same stack in your own cloud. See our security practices and subprocessors.
- Prompts and tool output
- Logs and alerts
- Source code and scan results
- Findings and reports
- Secrets seen during a run
- Your customers’ data
Questions
- No. We serve the same open-weight models, GLM-5.3 and DeepSeek V4.1 1017. The speed, cost and accuracy gains come from our inference infrastructure: how context is cached, compressed and routed.
- Agents for alert triage, incident investigation, vulnerability scanning and penetration testing. Anything that reads long logs, code or tool output over many steps benefits most.
- The open-weight models we serve run authorized security work, like exploit analysis and penetration testing, without the refusals common in closed models.
- Yes. Start on the Subconscious API, then move to a dedicated deployment in our cloud or on your own GPUs. The runtime is a drop-in replacement for vLLM or SGLang, and our engineers can install and tune it inside your cloud.
- The API serves GLM-5.3 Marathon and DeepSeek V4.1 1017 Marathon in both the OpenAI and Anthropic formats. Dedicated deployments can serve virtually any open model, including your own fine-tunes.
Run your security agents on Subconscious.
Start on the API. Move to a Dedicated Deployment in our cloud or on your own GPUs when you need it.