OpenAI announced on 2026-10-26 that CodeRabbit is using GPT-6.1 Sol as the primary model in its security agent, extending review beyond individual code changes to risks across complex codebases (OpenAI). CodeRabbit’s system also uses other frontier models and security tooling, according to the announcement (OpenAI).
Key Facts
| Item | Detail | Source |
|---|---|---|
| Model role | GPT-6.1 Sol is the primary model in CodeRabbit’s security agent | OpenAI |
| Scale | CodeRabbit reports flagging issues across thousands of repositories and organizations | OpenAI |
| Deep Scan | Investigation across repositories, including dependencies and how components interact | OpenAI |
| Remediation | For supported issues, a drafted pull request for a person to review and approve | OpenAI |
| Billing | Deep Scan is billed separately | CodeRabbit |
Timeline
- 2026-02-09: CodeRabbit described CodeRabbit Secure as security agents that trace exploitable paths across files, services and trust boundaries (Coderabbit (official statement)).
- 2026-10-26: OpenAI announced that CodeRabbit uses GPT-6.1 Sol as the primary model in its security agent (OpenAI).
Details (as Reported)
OpenAI explained that a pull request can affect shared libraries, dependencies and safeguards elsewhere in an application, and that GPT-6.1 Sol helps CodeRabbit connect behavior across files and components to surface potential security issues (OpenAI). During pull-request review, the model helps analyze changes as developers work (OpenAI). CodeRabbit Secure extends the investigation to linked repositories, such as a shared authentication library (OpenAI).
The announcement states that CodeRabbit’s review is independent of the agent that wrote the code and complements established scanners (OpenAI). For user-selected findings, CodeRabbit Secure can attempt exploit verification in an isolated sandbox without testing production systems (OpenAI). OpenAI added that human reviewers remain responsible for deciding which changes to ship (OpenAI).
CodeRabbit’s own page states that every finding goes through reachability, exploitability and blast radius analyses, and that findings no attacker can reach are capped at low severity (Coderabbit (official statement)). It also states that a verification pass rejects findings that are extremely hard to exploit (Coderabbit (official statement)). Fix with AI drafts a fix in a pull request, and a person on the customer’s team reviews and merges each fix, as CodeRabbit never merges one on its own (Coderabbit (official statement)).
On data handling, CodeRabbit states that it never trains models on customer code and shares it with OpenAI, Anthropic or both only for reviews (Coderabbit (official statement)). Cached data for private repositories is encrypted and expires within seven days, according to the company (Coderabbit (official statement)). PR Security Review, Blast Radius, Security Architecture Review and the Attack Surface Map come with the Advanced and current Enterprise plans, and new organizations get a 14-day Advanced trial (Coderabbit (official statement)).
What Is Not Yet Confirmed
The sources do not list open questions.