AI Models
RedCell runs on a curated catalog of models specifically benchmarked and tuned for authorized offensive testing, code audit, and threat hunting tasks.
Available Catalog
| Model ID | Context | Reasoning | Best For |
|---|---|---|---|
mimo-v2.6-prodefaultMiMo-V2.6-Pro | 200K | low · medium · high | General security reasoning, offensive and defensive. |
glm-5.3GLM 5.3 | 200K | low · medium · high | General security reasoning, offensive and defensive. |
deepseek-v4.1-flashDeepSeek V4.1 Flash | 1M | low · medium · high | Fast, low-cost runs and very long investigations. |
Reasoning Effort Settings
You can adjust how deeply the model thinks before producing a final turn or executing tool calls:
- low — Fastest response time, suitable for straightforward code syntax lookups or simple log parsing.
- medium — Balanced reasoning budget for code security reviews and general incident analysis (Default).
- high — Maximum deep thinking cycles for complex threat hunting, malware behavior analysis, and multi-file code auditing.
Switching Models per Execution
Pass the --model flag to select a model for a specific command run:
redcell exec --model
❯redcell exec --model glm-5.3 "Analyze telemetry dump in ./logs"
Using model: glm-5.3 (200K context)...