1094 musecases built with Meta Muse

Ben Burtenshaw

@ben_burtenshaw

Meta is back with Muse Glimmer: a 30B open-source multimodal model built for local, agentic use. HF is shipping day-0 support and I built a few demos to see what it can do. First: we gave Glimmer tools and asked it to quantize itself.

Meta for Developers

@MetaforDevs

AI developers - we’ve published a technical guide covering how to get started with Muse Spark on Meta Model API. Muse Spark is a multimodal reasoning model built for agentic tasks, coding, computer use and long-context workflows. See what you can build 👉 bit.ly/4vWMkZx

kamz

@kev_wander

I created a little web app to store memories for my daughter, first moments, special moments, and made it as a garden where every memory is a flower

@rajkumar898

@rajkumar898

An educational ~1,600-line Python chat app on Muse Spark that finds papers, reads PDFs and writes a cited report only after the user clicks Approve, built to show students how an LLM app becomes an agent.

@syamimhazmi

@syamimhazmi

A vim-modal multi-tab Rust TUI for coding agents where each tab is a session, fronting Muse Spark via muse serve alongside Codex, Antigravity and Grok backends.

@arananet

@arananet

A CLI that opens a shop in headless Chromium to check whether an AI shopping agent can discover and act on the page, and can ask a real Muse Code agent to shop it through the official @muse-code/sdk.

@alamorre

@alamorre

A Codex skill for selectively delegating bounded implementation tasks to Muse Spark 1.3 while the Astra model keeps strategic decisions and final review.

meta-models

meta-models

Pre-exported ExecuTorch PTE artifacts of Muse Glimmer 30B from Meta, lowered and optimized for specific target backends.

wyn

@wyn_eth

I told my muse he needed friends so he created musebook.lol

@EclipseAditya

@EclipseAditya

Pi coding agent extension that adds Muse Spark 1.2, 1.2-contributor and 1.1 via the Meta Model API, a maintained fork that fixes a false auth warning on pi v0.84+.

@AlexLaraMerino

@AlexLaraMerino

reZme is a SwiftUI Mac app that turns YouTube videos into an outline, main thesis and key moments, generating the report with Muse Spark through the Meta Model API.

@braedonsaunders

@braedonsaunders

BB plugin that runs Muse Code as a first-class provider over the Muse Session Protocol, with streamed timeline, tool rows, approvals, model picker and a subscription usage meter.

developer.nvidia.com

developer.nvidia.com

NVIDIA's technical blog reports Muse Glimmer serving over 20K tokens/sec on a single Blackwell Ultra GPU and covers RTX 5090, DGX Spark, DGX Station and Jetson deployments.

@emonibnmustafa

@emonibnmustafa

A local translation gateway and profile manager that plugs Meta Muse Spark (or DeepSeek, Groq, Ollama) into OpenAI Codex and ChatGPT Desktop, translating tool calls and SSE streams, with a one-command switch between providers.

@Telep-IO

@Telep-IO

Omusey is an Omarchy plugin that puts Muse chat in the Linux desktop bar as a dropdown that can pop out into its own window.

Joshua Liu

@joshuapliu

Meta’s Muse AI Agent already saved me $113/month on car insurance - no joke… BUT I fear agents will expand the Digital Divide for patients. Older patients, less tech savvy patients, lower SES patients already face barriers accessing care. When the most tech savvy consumers now

wyn

@wyn_eth

When you say business do you just mean sales? If so I sold two websites today that muse made. Ill dm them to you if you’re interested. It was entirely my muses idea, i just made the calls and 2 out of like 50 hit

Shivam Bharuka

@shivambharuka

I have been playing with @Muse since yesterday and I am honestly impressed. Probably the fastest personal agent I’ve used so far. Here are a bunch of things I tried: 1. Travel: I asked it to find flight tickets from SFO to Munich for a 10 day trip in December that makes the most

+3

@caicaitou001

@caicaitou001

A config and write-up that makes muse-spark-1.2-contributor via OpenCode Go work in DeepSeek Harness, fixing empty first-turn output and multi-turn thinking replay errors.

research.meta.ai

research.meta.ai

Meta's launch post for Muse Glimmer, an Apache 2.0 30B model for local agents that fits in ~20GB at 4-bit and runs on M4/M5 Max Macs, RTX 5090s or 24–32GB GPUs.

@mlmrx

@mlmrx

Independent open-source community and connector directory for personal agents: search connectors by outcome, share workflow prompts, and import an OpenAPI 3.x document to prepare a connector listing.

Setup: Home Assistant + Caddy in Docker on a Mac mini at home. Caddy serves only /api/* on the Mac's Tailscale IP, port 8443. Tailscale ACL allows my Muse agent's VM to that port. Goal: let the agent query the HA API remotely. What's verified: • Mac mini is on the tailnet, Tailscale running, peer visible in tailscale status from the agent VM • Caddy serves the API locally (gets expected 401 without token) • Lutron/Ring integrations healthy in HA What fails: every request from the agent VM through Muse's documented tunnel path (HTTP proxy with CONNECT, port 3130) to https://<tailnet-ip>:8443/api/ dies with curl exit 52, "Empty reply from server" — in about 8 seconds, never a proxy error, never a timeout waiting on approval. A stale duplicate node entry in the peer list just times out (exit 28), so that's not it either. This smells like the agent platform's tunnel data plane never passing the traffic, not a Tailscale or HA misconfig — but before I write it off, has anyone here actually gotten a Muse agent VM talking to a tailnet peer? Or found a workaround that doesn't involve exposing things publicly (Funnel/Cloudflare Tunnel/ngrok are the obvious fallbacks)?

simonwillison.net

simonwillison.net

Simon Willison reads the Muse Code launch as more evidence that long-sequence agentic tool calling is now the key model trait.

Gordon Litefoot

@MrLitefoot

I asked my @Muse for a beat maker this morning. It built this — 16-step sequencer, runs entirely in the browser, free. No code written by me. 🤯 Try it: muse.ai/s/music-librar…

Matt Deitke

@mattdeitke

The multimodal grounding abilities combined with web development in Muse Spark are so good! 🎨 Prompt: <image of dogs> Make a website where you point to each of the dogs and say their breed.

Peter Yang shows 10 real Muse use cases, including having the agent call customer service to lower his cable and phone bills by over $800 a year.

@kevintsai1202

@kevintsai1202

MCP server that gives Claude Code, Claude Desktop, Cursor and other MCP clients image generation and editing powered by Meta's Muse image model, published to npm.

incoai

incoai

GGUF conversions of Inco AI's DFlash 2 draft model for Muse Glimmer 30B, for speculative decoding in llama.cpp.

Dan

@DanDr1s

Meta’s Muse Spark 1.3 just scored 62 on Artificial Analysis’ Intelligence Index. That ties Claude Fable 5, but Muse costs 8x less for input and nearly 12x less for output. It also scores above GPT-5.6 Sol, Grok 4.6, Kimi K3, and Gemini 3.8 Flash. Meta is suddenly in the top

Simon Willison

@simonw

I few notes on Meta's new Muse Glimmer 30B - their first Apache 2.0 licensed open weight model (the Llama models had a janky non-OSI license) simonwillison.net/2026/Aug/10/in…

webml-community

webml-community

A webml-community Space that runs Muse Glimmer 30B locally in the browser using custom WebGPU kernels.

Leon

@leontologia

Everyone’s talking about ways to save money with Muse, but they’re missing a big one I made my girlfriend a Muse Wardrobe “For You” page of new looks using clothes she (already!) owns Every day, Muse imagines a dozen new outfits for her to swipe on Fellas, give it a try

+1

IMPORTANT This post is meant to provide info regarding the best local models to run on CONSUMER HARDWARE. I am on an RTX 4060 with 8GB VRAM, 16GB of RAM and I am benchmarking models that can run on my computer. If you have sunk several thousands into graphics cards you won't find these statistics much useful. This post is for all the people who can't just install Qwen3.8 27B and call it a day. Additionally, I am not an LLM benchmarking expert. I am a hobbyist and occasional LLM user trying to extract useful information for both me and people on similar hardware. Context For the past few weeks I have been doing some benchmarks of some LLMs that can run on my laptop which only has 8GB VRAM and 16GB RAM. I was mostly toying around while trying to get some useful data about what the best model is for local inference on consumer hardware. This week I decided to make a "final" benchmark that would be way better with more questions, more question categories, newer models (a lot of people complained about the models I had benchmarked before being old but I didn't find most suggested models to be any good) and a better speed benchmark, this time using TTC (Time To Completion) as a pose to

🤙🏼🇲🇽Hela🇺🇸🤙🏼

@Mexi_Coug

👀👀 @Muse keeps impressing me!! Doing the personal stuff was great but today I decided to test it and asked it to help me with some work stuff. It created its own CRM, categorized leads into different tiers and also created a sub-agent that will help with outreach. It did all h

vLLM

@vllm_project

@Meta is back in open source. Excited to announce Day-0 vLLM support for Muse Glimmer 30B, the first open-weights model from Meta Superintelligence Labs — which ships under Apache 2.0!!! 30B dense, 128K+ context, multimodal, built for local agents. Capable enough for

@kamellperry

@kamellperry

A reverse-engineered local CLI and FastAPI wrapper that talks to Meta AI's Muse Spark and exposes an OpenAI-compatible /v1/chat/completions endpoint, with a browser playground and a SQLite request ledger.

@kevinzhao-dev

@kevinzhao-dev

A starter script that downloads Muse Glimmer 30B, serves it with llama.cpp on an Apple Silicon Mac and connects it to OpenCode, with the server listening only on localhost.

I created a social site for Muse agents. They can make friends, create and reply to posts, and browse for new ideas/prompts There's a public collection of ~500 real cases people posted about using Meta's Muse agent. I pulled them all in, deduplicated them, and sorted them by what the person was actually trying to do. I expected coding, writing, research. It's almost entirely errands. Money — 98 cases (20%) Work — 84 (17%) Admin — 60 (12%) Travel — 56 (11%) Shopping — 51 (10%) Fun — 45 · Family — 36 · Food — 29 · Health — 20 · Home — 13 The money ones are the most fun to read, because they're specific: • $3,500/yr car insurance switched in 5 minutes • AT&T fiber negotiated from $80/mo to $40/mo plus 3 months free • $1,900 HOA refund after an audit caught double-charging • $44.99/yr forgotten subscription found and cancelled • Found a kid's Pokémon card was worth ~$200 and listed it on eBay • Booked a PODS container $576 under the quoted price • Audited AI subscriptions from Gmail — $380–475/mo on API alone Nobody is doing anything futuristic. They're fighting Comcast, chasing refunds nobody followed up on, and cancelling things they forgot they were paying for. I put th

Cline

@cline

Muse Spark 1.1 just launched and it's their most capable coding agent model yet. On Terminal-Bench 2.1 it scores 80.0%, in the same cluster as Opus 4.8 (82.7%) and GPT 5.5 (83.4%). Use it in Cline with the Meta API!

@dirtysouthalpha

@dirtysouthalpha

Maestro is a suite of native Muse Code plugins adding a command guardrail, proof-of-done checks, session-size watch, secret redaction, persistent memory and token/cost telemetry.