Thread

Meta's Muse appears to use an OpenAI model labeled muse-special

143 points · 51 comments · Aeroi

Aeroi · 1790360286 · focus · HN ↗
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: <a href="https://x.com/heypeterjames/status/2103545183400800746" rel="nofollow">https://x.com/heypeterjames/status/2103545183400800746
pete at mouse dot dev
1. haolez · 1790362252 · focus · HN ↗
  Nice work!
  Are you sure this isn't just a custom model that is API-compatible with OpenAI?
  1. vineyardmike · 1790367411 · focus · HN ↗
    This feels like the far more likely option.
    The Muse public APIs seem to be heavily inspired by OpenAI's, they even support the richer "responses" API. They have a similarly shaped compaction API, they have the same "encrypted" reasoning. Even the Muse harness seems heavily inspired by (open source) Codex. Considering Azure's historic involvement w/ OpenAI, it seems more plausible that Azure is used as spill-over capacity, and the same Azure infra used to encrypt GPT models encrypts Muse models.
    My guess that the generally visible Anthropic/OpenAI details are leftovers from the whole meta "move fast" behavior. There are a few rough edges on the product where it leaks internal codenames (eg. signing in w/ WhatsApp required consenting to using "hatch" on one screen), so I'd believe that this was a leftover on the VM from the prototyping stage, before the Muse models were ready, or because the dev's got to benchmark against different models.
    jychang · 1790421106 · focus · HN ↗
    > The Muse public APIs seem to be heavily inspired by OpenAI's, they even support the richer "responses" API.
    ???
    Every single provider basically copy openai's api endpoint
    <a href="https://api.openai.com/v1/chat/completions" rel="nofollow">https://api.openai.com/v1/chat/completions
    <a href="https://router.huggingface.co/v1/chat/completions" rel="nofollow">https://router.huggingface.co/v1/chat/completions
    <a href="https://integrate.api.nvidia.com/v1/chat/completions" rel="nofollow">https://integrate.api.nvidia.com/v1/chat/completions
    <a href="https://generativelanguage.googleapis.com/v1beta/openai/chat/completions" rel="nofollow">https://generativelanguage.googleapis.com/v1beta/openai/chat...
    <a href="https://api.x.ai/v1/chat/completions" rel="nofollow">https://api.x.ai/v1/chat/completions
    <a href="https://api.deepseek.com/chat/completions" rel="nofollow">https://api.deepseek.com/chat/completions
    <a href="https://openrouter.ai/api/v1/chat/completions" rel="nofollow">https://openrouter.ai/api/v1/chat/completions
    <a href="https://opencode.ai/zen/v1/chat/completions">https://opencode.ai/zen/v1/chat/completions
    <a href="https://api.cerebras.ai/v1/chat/completions" rel="nofollow">https://api.cerebras.ai/v1/chat/completions
    Saying "Muse's public APIs seem heavily inspired by OpenAI" is like saying "a web server’s API seems heavily inspired by HTTP". OpenAI’s API format has become an industry standard.
    vineyardmike · 1790454999 · focus · HN ↗
    Thanks for the links to the “completions” APIs that are common. Not the “responses” API I’m talking about, however.

Open on Hacker News to reply ↗

Unofficial Hacker News client; not affiliated with Y Combinator.