Theme 1 – Muse looks like an OpenAI‑compatible model running on Azure
- “If it's using the same weird (and awful) 'backend prompt encryption' pattern and mechanism that Codex also uses … then I'd also say it points to OpenAI being involved.” – embedding-shape
- “Its very openai shaped and served differently from all the other models listed in the daemon, under a mysterious name.” – Aeroi
- “The Muse public APIs seem to be heavily inspired by OpenAI's, they even support the richer 'responses' API.” – vineyardmike
- “It uses a similar API to openAI. Source: I work on AI at Meta.” – VygmraMGVl
Theme 2 – Why Meta would host Muse on Azure (capacity spill‑over, partnership, or overflow)
- “GPU compute capacity?” – losteric
- “Microsoft owns 27% of OpenAI.” – xmcp123
- “If this is a model produced by Meta with an API designed to be compatible with OpenAI models, but it's not actually an OpenAI model, why is Meta hosting it on Azure?” – thinkling
- “Meta already serves its own models on Azure under their real names … those sessions do not come back as gpt_responses_v1 items.” – Aeroi
- “Because there was an article written in Bloomberg about how they're looking to sell their extra compute.” – fg137
Theme 3 – Strange outputs (language switches, odd phrases) hint at training‑data artifacts or leftover debug prompts
- “Muse 1.3 spark occasionally spits out Chinese character responses to me like internal instructions. ‘Go fast’ or ‘Get help’…” – manav
- “I had a totally benign chat with OpenAI and it titled it as ‘amateur porn’ in Chinese characters …” – wincy
- “I remember getting a bunch of ‘Thanks for watching! Subscribe and smash that like button’ in the middle of chat sessions …” – dgellow
- “If you ever look at what's in the Common Crawl the amount of Chinese porn crap is jarring.” – fg137
- “People that know multiple languages sometimes code switch …” – tehjoker (suggesting model may be reflecting mixed‑language training data).