Meta's Muse appears to use an OpenAI model labeled muse-special
After my previous post on Muse's runtime export hit Hacker News, I dug deeper into the filesystem and found a single session using a model labeled azure/muse-special. The logs show OpenAI-style call IDs and an encrypted gpt_responses_v1 signature, suggesting it's a GPT model served through Azure. The model catalogue also includes Claude and GPT variants, and there's Anthropic client code with a kill switch. I explore whether Meta is distilling from OpenAI or Anthropic, but the encrypted reasoning and RL server restrictions suggest not. This is my own exploration of a fascinating release.
The raw chain of thought is encrypted and the RL server refuses those blobs. There is no indication that Meta copies OpenAI or Anthropic weights.
- Aeroi
author here. I kept digging through the Muse filesystem after my first post hit the front page here this week.
Going through my logs I found a background agent that used a model called azure/muse-special while building my website. I found this interesting and dug a little deeper.
The transcript and daemon binary point to an OpenAI model running on Azure. Still unclear which one or why it was selected.
The runtime also ships with an Anthropic client and a catalogue listing Claude, GPT and Kimi models. I didn’t observe Claude being used in my own sessions, but this is my Part 2 of exploring the Muse file system.
original post: https://x.com/heypeterjames/status/2103545183400800746
pete at mouse dot dev
- Tiberium
I don't really see evidence that it's an "OpenAI" model, the title is IMO very misleading.
- 6thbit
It'd be hilarious that meta chose to pay openai. If there was such a deal, would it come to light in any public/official filings?
- manav
Muse 1.3 spark occasionally spits out Chinese character responses to me like internal instructions. “Go fast” or “Get help”…
- GenerWork
I'm not sure I understand why they would route it to other models. It can't be that they don't have enough compute. Maybe worried that the answer from their own models would be bad? Doesn't really make sense, but I could be missing something.