1076 musecases built now with Muse

Hi all, Profile v2.2 is out. It's an open-source optimizer for inference servers. It computes your GPU's roofline ceiling, measures your live server against it, names the bottleneck, gives the flag. You apply. It re-measures. Every fix answers to a number. vLLM only today. More engines next. This release: core rule engine rewritten. Eight rules on a priority DAG with mutual exclusivity. Five alarms fire, four echoes are silenced, one true cause survives. Deterministic. AMD cards are now supported too. Tuning today is chaos: try a flag, wait, squint at a dashboard, repeat for days. Profile turns it into deterministic engineering: measure, fix, verify. Results in a few iterations. Mine took 4, ~30 minutes. My setup: RTX 5090, muse-glimmer 30B, SWE-Bench agents, no DFlash spec decoding. • 81 → 421 tok/s at 25k ctx • $3.41 → $0.65 per 1M output tok • TTFT 224ms (p95 500ms), TPOT 23ms at end of run • 4.72 → 1.08 J/tok https://preview.redd.it/4vazyxkcq6kh1.png?width=2248&format=png&auto=webp&s=77923a489b6f725240d23a7953150b5779260734 One iteration regressed hard: KV thrashing, TTFT 32.8s. Profile labeled it worse. Next fix recovered it. Regressions stay in the record. Watc

Vercel added Muse Spark 1.3 (meta/muse-spark-1.3) to AI Gateway on launch day, with 1M-token context, text/image/PDF input, and both standard and contributor pricing tiers.

@NeedsChloesure

@NeedsChloesure

Muse Proxy is a Cloudflare Worker gateway that stores CalDAV/CardDAV credentials once and issues scoped API keys, so the Muse agent can reach self-hosted calendars without holding passwords.

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.

Shabbir

@shabzcohelp

Had it check for a slot for a swimming class for both my kids. Took 2 days, but found one and booked it for me without me having to continue refreshing

Matrix Ɪnvesting

@Matrix_invests

Super fun fact I just found about @Muse!! If you connect your Robinhood banking account , it’ll be able to see your investment portfolio and provide insights and create finance goals and plans Sorta a loophole for a Robinhood connector, super cool!! @alexandr_wang @MattPRD

@o1ie12

@o1ie12

A single-file marketing site (hash routing, GSAP transitions, Lenis scroll, Three.js) built by Muse Spark in Meta AI from 5–6 prompts, with no manual code edits.

@saipreetham99

@saipreetham99

MdChat is a native macOS Markdown previewer and vim-keyed editor with an AI side panel that answers questions about selected blocks and rewrites them behind a diff, using Gemini or Muse Spark.

Meta's developer portal for Muse connectors: describe a connector, submit it for functional, security and legal review plus end-to-end testing, and get listed in the Muse directory.

AI John walks through installing OpenCode, creating an API key and connecting Muse Spark 1.3 on the Contributor Free tier, noting prompts may be used for training.

We tried using Meta's new Muse Code agent, but it has a bug that doesn't let it sign in from a docker container. So we did a fun experiment: Meta claims Muse Spark 1.2 was co-trained with their Muse agent harness. So we extracted instructions from their system prompt and added them to the Cline harness. TL;DR of this special prompting: - Trust source code over the user prompt, so read every call site and existing tests before starting the task - Weigh edge and error cases as heavily as the happy path - Always reproduce the bug before fixing - Don't trust the first passing test suite, and verify suspicious looking half-baked tests - Never stop at just editing, keep working until the change is verified complete. We then asked this modified harness to fix a real bug from our repo, and compared the results to the original Cline agent harness. Results: - Used 2.7x fewer tokens (19.7M → 7.2M) - Finished 2x faster (49min → 24min) - Cost 2.4x less ($7.69 → $3.25) Same Muse Spark 1.2 model, same task, only the prompting changed. Incredible how much of a performance gain Meta was able to achieve training it on these special instructions!

@carlosrymer

@carlosrymer

A browser-only agent console for Muse Glimmer 30B that renders the model's ATEM tool-call format client-side, streams from a local server and runs every tool inside the tab across seven capability tracks.

@shelajev

@shelajev

A Docker Sandboxes kit that runs Meta's Muse Code CLI in an isolated sandbox, allowing network access only to the Meta services needed for install, auth and model requests.

John Kennedy

@CommerceJohn

Just booked flights on @Muse. Amazing experience. Infinitely better than @Expedia who I have been using consistently for years.

Michael

@OffZeroCyber

Here is how I started the conversation with my @muse: I want to experiment with turning you into a personal email security agent for my Gmail. The goal is to see how far you can go toward acting like a lightweight ICES/email security platform for an individual user. Start by

AICodeKing reviews Muse Glimmer for local agent setups, finding it strong at tool calling, multi-step tasks and failure recovery but weaker on general benchmarks.

Jacob Mitchell • FermentIQ

@jacobdmitch

Yes! I’ve already connected mine to @getFermentIQ @render @ElevenLabs @TransistorFM @github and more!

anubhav

@erfahrung

@Muse found me a CPA, handled all communication and locked me in for next year. The kind of help I always sought!

Patrick Donohoe

@patrickdonohoe

Introducing floo, a cloud platform where your personal agent can build and run apps for you. Tell Muse what you need. It builds the software. floo handles hosting, data, and login, giving your Muse tools it can keep using and you an app to see and guide its work. Launching

Unsloth's docs page for running Meta's Muse Glimmer 30B locally with its GGUF quants.

Abdul

@AbdulTheBuilder

Big fan of @Muse Took grok bot & ran with it, haven't tried any serious work with it yet but it's working on getting me a refund from enterprise car rentals, booked me a flight for a Christmas trip & found me some pretty sick deals on recliners all within less than an hour of

Marcus

@marcustayye

Meta employee here! I built a finance calculator for Singaporeans to help with adulting decisions, like mortgages, retirement, savings, and a bunch more! Entirely with Muse! gethowmuch.net

DFlash2 drafter for Muse Glimmer 30B fine-tuned on long-horizon, on-policy software-engineering traces to speed up coding workloads.

Venelin Valkov pairs Muse Glimmer with Hermes Agent on llama.cpp for a fully free local agent, testing whether a better harness fixes the model's mixed early reviews.

Artificial Analysis puts Muse Spark 1.1 at 51 on its Intelligence Index, 8 points above 1.0, and calls it cost and token efficient versus peers.

@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.

OpenRouter

@OpenRouter

Muse Spark 1.1 is available now on OpenRouter for US-based developers. openrouter.ai/meta/muse-spar…

Setup the latest (master) version of llama.cpp server with the guide and the official GGUF weights by Meta AI: https://huggingface.co/meta-models/Muse-Glimmer-30B-GGUF and connected the Hermes Agent to the llama.cpp endpoint. Getting about 22t/s (+3-4t/s) on M5 Pro, using ~24GB including the drafter (provided by Meta). The model did correct tool calls and actually did some useful work inside the Hermes Agent. Moreover, the resulting coding task/project works, which was not the case when running the model with OpenCode. Watch more: https://www.youtube.com/watch?v=cmENEolUtM4

Fei Xia

@xf1280

Muse Spark is agentic, which means you can ask it to leverage different test-compute scaling methods to improve quality. Here I ask the model to use parallel subagents to do counting and the results are greatly improved! meta.ai/share/aD4KAPeV…

Theagiledev

@matthewjamesr

Hopefully we gain traction with #AgileFactor. A muse-enhance global conflict podcast with daily episodes release across YouTube, Spotify, and Apple Podcasts. Show + Interactive episode dashboards! matthewreichardt.com/agilefactor/ youtube.com/@TheAgileFactor

No Saber Ni Papa

@NoSaberNiPapa

@Muse renegotiated our AT&T internet bill today. From $80/mo to $30/mo w/ 3 mo free & they 2X our speed too. If you want to try Muse, it's free. Use my code **5QD8CO** in Settings within 48 hours of signing up & we both get 1 billion bonus tokens. muse.ai/join

Jonny Caiani

@caiani_jonny

I’m using Meta Muse to book a three day trip and it’s a breeze! A little slow doing the tasks sometimes but hey it’s better then me doing the tasks

Artificial Analysis' independent benchmark write-up of Muse Spark 1.3.

Valiant Labs' Esper 4 agentic coding fine-tune of Muse Glimmer 30B.

I'm like you guys and am constantly experimenting with new models, seeing what they're all good at, how I can make use of them for certain projects and goals. I've been using Qwen 3.8 27b for minor coding work and it has been impressive. But with just regular chatting I have been impressed with Muse Glimmer. It seems to be able to have the ability to follow and hold good, deep and meaningful conversations without coming off as a typical chatbot. No repeated statements like "I hear what you're saying", "that sounds really deep..." none of what sounds generic or like it's blowing smoke up your ass. I was impressed with how natural it comes across just in natural conversation. I think it's one of the best "chat" models you could get right now as it's one of the only local models that doesn't feel like you're chatting with an AI when having a conversation. I'm thinking of finding a way to run both Qwen3.8 and Muse at the same time. It's fun to play with these things.

james

@ads4apps

Having @Muse find the best offer to sell my Model X. Then it’s finding a lightly used new version of the same car

Kingy AI explains Muse Glimmer's benchmarks, 24-64 GB hardware needs, GGUF setup, pricing, runtimes and risks.

Independent benchmarks and analysis of Meta's open-weight Muse Glimmer.

I made a 400 page pdf html docs words txt from 400 Google docs any ai/agent app can't make it I Tried

@everyai-com

@everyai-com

A set of small fixes that let Muse Spark run in pi, Agent Orchestrator, opencode and other OpenAI-compatible tools, handling its non-standard SSE event and missing model-catalog entries.