GoKawiilLeaked logs suggest Meta's Muse AI routes some tasks through an Azure-hosted OpenAI model
A researcher digging through Meta's Muse filesystem, following an earlier Hacker News-featured investigation, found that nearly all session logs used Meta's internal 'Avocado' model, but one subagent session was routed to a model labeled azure/muse-special. Technical markers in that session—an OpenAI-style tool call ID format, a gpt_responses_v1 signature, and an encrypted payload prefix associated with OpenAI—suggest the model may be a GPT model served via Azure's OpenAI service, though the exact model version and reason for its use remain unconfirmed.
GoKawiil's interpretation of the reporting above, not reported fact.
If confirmed, this would indicate Meta's Muse product supplements its own in-house model with third-party AI infrastructure from a competitor, which could raise questions about dependency, cost, and how Meta represents its AI stack to users. The researcher's evidence is circumstantial—based on naming conventions and formatting artifacts rather than direct confirmation from Meta or OpenAI—so the claim remains an inference pending official comment.
- A researcher found a model called 'azure/muse-special' used in one Muse subagent session, distinct from the usual 'Avocado' model.
- Technical signatures in the session (OpenAI-style call IDs, gpt_responses_v1 tag, gAAAAA-prefixed encrypted payload) hint the model may be an OpenAI GPT model accessed via Azure.
- Neither Meta nor OpenAI has confirmed this, and the exact model identity and purpose of its use remain unclear.
Source: mouse.dev, 2026-09-25
Published there as: “Meta's Muse appears to use an OpenAI model labeled muse-special”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.