AI Models
Cyrion AI runs on the latest frontier and open models. Choose the model for each scan, or set a default in Settings → LLM. The models you can select depend on your plan.
Cyrion AI runs on the latest frontier and open models. Choose the model for each scan, or set a default in Settings → LLM. The models you can select depend on your plan.
Latest models
| Provider | Models | Context window |
|---|---|---|
| OpenAI | GPT-6 Astra, GPT-5.6 Sol, GPT-5.6 Terra, GPT-5.6 Luna | 1M |
| Anthropic | Claude Fable 5.1, Claude Opus 5, Claude Sonnet 5 | 1M |
| Gemini 3.8 Flash, Gemini 3.1 Pro, Gemini 3.1 Flash Lite | 1M | |
| DeepSeek | DeepSeek V4.1 Flash, DeepSeek V4 Pro | 1M |
| xAI | Grok 4.6, Grok 4.20 | 500K to 2M |
| Qwen | Qwen3.8 Max, Qwen3.8 Flash | 1M |
| Moonshot AI | Kimi K3 | 1M |
| Z.ai | GLM-5.3, GLM-5.3 Flash | 1M |
| MiniMax | MiniMax-M3 | 1M |
| Open and free models | Nemotron 3 Ultra, Gemma 4, Nex N2.5 | up to 1M |
Long context windows let agents hold an entire scan in view, so they stay coherent across thousands of steps.
What each plan includes
| Plan | Model access |
|---|---|
| Free | Free open models, for the demo scan. |
| Lite | Free models plus a curated menu of fast, cost-efficient models: DeepSeek, Qwen and Gemma. |
| Hacker | The full standard catalogue plus frontier flagships such as Claude Opus 5, Claude Fable 5.1, GPT-6 Astra and Grok 4.6. |
| Red Team | Everything in Hacker, plus uncensored models built for AI red teaming. |
| Enterprise | Everything above, plus custom models. |
Info: A model you cannot yet use is shown with the plan that unlocks it, so you can see what an upgrade adds.
Defaults
| Plan | Default model |
|---|---|
| Free and Lite | DeepSeek Flash |
| Hacker | Claude Sonnet 5 |
| Red Team and Enterprise | Claude Opus 5 |
Model tiers inside a scan
Cyrion AI matches the model to the job so you do not pay flagship rates for routine work.
| Tier | Used for |
|---|---|
| Large | Orchestration, vulnerability analysis and exploitation, where deep reasoning matters. |
| Medium | Reconnaissance, analysis and sub-agent tasks. |
| Small | Summaries, data extraction, note taking and report generation. |
By default one model serves every tier. In Settings → LLM you can assign a different model to each tier, for example a flagship for Large and a fast model for Small.
Choosing a model
- Fast models such as DeepSeek V4.1 Flash, Gemini 3.8 Flash and Qwen3.8 Flash suit Quick and Standard scans where cost matters most.
- Balanced flagships such as Claude Sonnet 5 and GPT-5.6 Terra are strong all-round choices.
- Maximum reasoning with Claude Opus 5, Claude Fable 5.1 or GPT-6 Astra suits Aggressive and Thorough scans on complex applications.
Models consume credits in proportion to the work they do, so flagships draw down your balance faster. See Credits & Usage.