Use o3 (openai/o3).
This is one of the cases where paying for more reasoning is justified.
~8.6× more per run.
Which AI model should you use to review a pull request for security?
Why ->
The task description signals deeper reasoning, trade-off analysis, or risk-heavy judgement. A larger tier earns its cost here because it handles longer reasoning chains and ambiguity more reliably.
Cost breakdown
| Model | Provider | Tier | Input cost | Output cost | Est. / 1k calls |
|---|---|---|---|---|---|
| o3 openai/o3 | Openai | Deep | $0.002000 | $0.008000 | $10.80 |
| GPT-5 Mini openai/gpt-5-mini | Openai | Moderate | $0.000250 | $0.002000 | $1.25 |
Matched signals
- review
This recommendation is based on the task description alone. If you are attaching a large codebase, mention it in the input above.
Model pricing: OpenRouter model catalog, retrieved April 17, 2026. Classification: rightmodel ruleset v2.
Related tasks
Review SQL Injection Risk in Code
DeepInspect a query path for interpolation, unsafe ORM usage, and missing parameterization safeguards.
Use o3.
Audit an Authentication Flow
DeepInspect an authentication flow for trust boundaries, session handling, and privilege escalation risk.
Use o3.
Summarize a Pull Request
RoutineCondense a pull request into key changes, risks, and reviewer notes for fast handoff.
Use Gemini 2.0 Flash.