Inquiry About the LLMs Used by Rovo

[JCN_ToM運用]

Hello,

Could you please provide information on all Large Language Models (LLMs) currently used by Rovo, including their specific versions (for example, GPT-4o, Claude Sonnet 4, Gemini 2.5 Pro, etc.)?

If this information is not publicly disclosed, we would appreciate it if you could kindly confirm that the details are confidential or not publicly available.

Thank you for your assistance.

Best regards,

[Sami Shaik]

Hi @JCN_ToM運用 , good news: this is partially disclosed, and I can point you at the official written sources your assessment can cite.

What Atlassian discloses: the model families, not specific versions. Atlassian's official Rovo data usage and privacy page states that Rovo uses OpenAI LLMs (GPT), open-source models (Mistral, LLaMA), and third-party hosted LLMs (Claude, Gemini) across Rovo Search, Chat and Agents. Reference: https://www.atlassian.com/software/rovo/guides/admin-guide/rovo-data-usage-privacy

What is not published: exact model versions. Rovo routes requests dynamically across models per task, the model is not user-selectable in Search, Chat or Agents, and Atlassian does not publish a version-level list (for example which GPT or Claude version serves a given feature). So for your records: the families are public, the specific versions are not publicly documented, and that is by design of the routing architecture rather than an omission. The per-feature disclosures live in the AI Transparency Notes: https://www.atlassian.com/trust/ai/transparency

One exception worth knowing: in Rovo Dev (the coding agent), the underlying model IS user-selectable in the CLI and IDE, currently across Anthropic Claude and OpenAI GPT options, so that is the one Rovo product where you see concrete model choices.

If the question behind the question is data flow: two facts for your compliance file. First, Atlassian's commitments state that LLM providers do not store your inputs and outputs and do not train on them. Second, there is an org-level Atlassian-hosted LLMs option where Rovo features rely only on models hosted within the Atlassian Cloud boundary (OpenAI and Meta open-source models, US-hosted), and prompts are then not sent to external LLM providers at all: https://support.atlassian.com/organization-administration/docs/atlassian-hosted-llms/

For a formal written confirmation of confidentiality beyond these public pages, your Atlassian account team or a support ticket is the right channel, since this forum is peer support. But the three links above are the official citable record, and they should cover most vendor-assessment requirements.