Entries matching “test”

Ported a Miniature Golf prototype to VR using Meta Muse Code and the Unity CLI, added dozens of automated tests, and built a website to capture test runs, screenshots and results.

Unsloth AI

@UnslothAI

2-bit Muse Glimmer GGUF managed to call 100+ tools on just 14GB RAM. 🔥 Muse Glimmer did a complete repo bug hunt for 5 mins nonstop with: evidence, repro, fix, tests and a PR writeup. Run and train it in Unsloth. GitHub repo: github.com/unslothai/unsl…

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.

DataCamp's Josep Ferrer ran Muse Spark 1.3 on three real coding tasks. Two used 23–32% fewer completion tokens, but a refactor used 70% more, for a net 12% cost increase.

My fun weekend project was to try to make the new Muse Glimmer 30B work with a longer context, deciding to go for 512k first. I had expected the usual YaRN shenanigans and maybe a LoRA. I couldn't have been wrong more. Upon closer look, Glimmer turned out to be rather unusual architecturally. The thing that make long-context adaptations painful in other models, full attention layers with token position encoding, it simply not there. Instead, only 2048 tokens-wide SWA layers have RoPE, and full GQA attention layers have no position encoding at all. It appears the model is trained to work with long-distance token relationships inferred from the context and SWA layers. It's a rather bold architecture bet, but it seems Meta managed to pull it off. As a result, the model architecture appears to be uniquely suited for context extension by simple mechanical means. To change model context length from stock 128k to, say, 512k, you need only to change “max_position_embeddings” config setting from 131072 to 524288. What confuses other models, like Qwen3.5 family, Glimmer just takes into its stride. I spent close to 70h of compute on DGX Spark to test stock model with extended context on a

Venelin Valkov runs Muse Glimmer 30B locally via llama.cpp server and tests it on coding with OpenCode, agentic tasks and frontend work.

kwindla

@kwindla

Exciting to see Meta releasing new open weights this week. Meta trained the new 30B dense model Muse Glimmer with "agentic" use cases in mind. This generally means task-oriented, multi-turn, and heavy use of tool calling. I ran Muse Glimmer in the GGUF quant through a bunch of

Avi Flombaum

@aviflombaum

Zuck and @Meta really crushed how amazing @Muse is. I told it about my parking app because I tried using it originally to find parking but it couldn’t do it and like 12 hrs later I got this alert that it could help me

STARTUP HAKK tests Muse Spark 1.3 and argues the harness, tools and context around a model matter as much as the model, pitching their OpenMonoAgent harness.

I have a classic test for local LLM's. I asked for 8 ball pool game with only one HTML file and Muse Glimmer spend 21k Token(I m using full context so 128k) and only created a 220 lines of HTML and said its done. With my experience its not even close to Qwen 3.6 27B and we are waiting for Qwen 3.8 27B already. What is your toughts about this model. I was so hopeful until this test.

Anyone using Muse an orchestrator for fleets of coding agents? I’ve found it surprisingly effective and it’s convenient to be able to do so on the go from my phone. Plus Muse can instantly serve your web app for quick testing. it’s actually been pretty fun.

Quick update on the RTX 5060 Ti local LLM repo. It has changed quite a bit since my previous posts. The project started as a collection of practical notes and benchmark results. That was useful, but as the dataset grew it became harder to answer the question most people actually had: What configuration should I run? I have rebuilt the repo around tested, copyable presets rather than treating every successful benchmark request as a front-page result. What changed? The project now separates three things: • Presets: exact configurations intended for people to copy and run. • Evidence bundles: reviewed proof of context fit, retrieval, sustained generation and performance. • Raw receipts: retries, failed experiments and diagnostic runs that are kept separate as engineering material without automatically becoming recommendations. The website now leads with the published preset catalogue. The larger results explorer is still there for comparisons and historical data, but it is no longer the first thing visitors have to decipher. There are currently seven published presets across the 1× and 2× RTX 5060 Ti lanes: 1× RTX 5060 Ti 16GB • Qwen3.8 27B IQ3_XXS at 64K with q8 K

bilbooo

@bilbooo__

Told Muse to find me some Hondas on FB Marketplace from 2010–2015 as a test, and it worked pretty well. Meta did a great implementation.

Muse Glimmer 30B quantized with Intel AutoRound at a 3.5-bit target and packed with llm-compressor, tested on vLLM.

LM Studio

@lmstudio

Muse Glimmer 30B is live in LM Studio! It's a new open source model from Meta. Apache 2.0 license, fit right on your laptop. It is the strongest model of its size class we've tested.

HolaClaw tested Muse Glimmer on base M3 and M4 Macs: it fits in 24GB of RAM but generates at 4.3 tokens per second.

Stark Insider introduced Muse to its own AI agent setup to test whether Muse's one-minute onboarding costs depth compared with OpenClaw.

I'm just using OpenWebUI with a simple FastMCP server. Every other model I've tried will simply run a few lines of Python and give me the result. Glimmer seems to overthink like crazy to the point of being useless. On the carwash test it tried to compute emissions using Python. I'm using the recommended sampling parameters, default template, and I've tried both unsloth's Q6_K_XL and Meta's dynamic GGUFs. Any ideas? EDIT: It seems like it's definitely related to the tools available. With them disabled, it's reasonably efficient. I guess it's just overly eager to call every tool it can unlike Qwen or Gemma in my experience.

@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.

AICodeKing tests Meta's Muse Spark 1.2 and the Muse Code terminal agent on the free tier, covering pricing and features before running KingBench 3.

ThursdAI hosts compare daily-driver assistants, then David Pawlan, who ran 273 tests across 23 assistants for Assistant Benchmark, walks through how Muse stacks up.

merve

@mervenoyann

Meta released Muse Glimmer 30B: multimodal model for your Claw/Pi setups 🔥 we tested and fine-tuned the model for you, and shipped day-0 support in transformers and llama.cpp, including DFlash for 2-4x speed-ups 🥵 read our blog huggingface.co/blog/muse-glim…

Simon Willison ran an 18.16GB build of Muse Glimmer locally, testing code exploration and image description. He found multi-step reasoning and tool use strong and creative generation mixed.

Alex Volkov tests the Muse agent's research and Stripe-powered payments, and walks through its VM security, personal-data handling and the training opt-out.

Simon Willison

@simonw

Pelicans for Meta's new Muse Spark models - plus I did a bit of a deep dive into the Code Interpreter and fascinating "container.visual_grounding" tools in their meta.ai chat UI simonwillison.net/2026/Apr/8/mus…

WTF Code uses Muse Spark 1.3 Contributor Free through OpenCode to build an interactive project from scratch, testing UI, animation, debugging and multi-step development.

Theo (t3.gg) tests Meta's Muse Code terminal agent powered by Muse Spark 1.2 and focuses on how cheap it is compared with Claude Code.

@newjordan

@newjordan

A tested Muse Glimmer 30B Q4_K_M serving package for Intel Arc Pro B70 with full 131k context, reporting 19.0 tok/s decode at 129k cached and 503 tok/s full-context prefill.

WTF Code shows how to wire Muse Spark 1.2 into OpenCode through the OpenCode Zen API and tests it as an Opus 4.8 challenger.

Bijan Bowen's first look at Meta's Muse Code terminal agent and Muse Spark 1.2, testing a browser OS, C++ skate game, CAD design, flight sim and subway FPS.

Rohan Paul

@rohanpaul_ai

Meta's open-sourced Muse Glimmer 30B vs Muse Spark 1.2 Glimmer delivered 5 runnable games at roughly one-fifth Muse Spark's cost. Very Interesting experiment by @aimlapi . the setup: • Muse Glimmer 30B — via aimlapi[.]com. cost: $0.02 • Muse Spark 1.2 — via aimlapi[.]com.

Cloud Codes runs Unsloth's 2-bit dynamic quant of Muse Glimmer 30B in about 14 GB of laptop memory and tests 100+ autonomous tool calls.

@mpfaffenberger

@mpfaffenberger

A reproducible single-GPU deployment of Muse Glimmer 30B in BF16 with DFlash speculative decoding on a 96GB RTX PRO 6000 Blackwell, served via vLLM with pinned overlays and smoke tests.

It's free to use, of course with meta you're probably the product if it's free but generally it's great! I set it up today and tested it a bunch. Hoping for some feedback, and if you need some help setting it up feel free to ask on here or let me know on the discord server! -Rob

AgentMail's guide: tell Muse to get an AgentMail address, paste an API key into its Secrets tab, approve the connection and send a test; about five minutes, no code.

KGP Talkie asked Qwen 3.8 27B, Muse Glimmer 30B and Gemma 4 26B the same twelve questions ten times each to see which answers the same way twice.

IchBinFabian installs Muse Code, compares Muse Spark 1.3 pricing with Claude Code and Codex, reviews the benchmarks and builds two real projects in OpenCode for a combined 82 cents.

OpenTools maintains a source-backed directory of about 163 Muse connector setup guides, separating Meta-documented, partner-documented and community skills (e.g. HubSpot, Attio, Ashby, Plaid, Slack), with status labels such as draft or maintainer-tested.

Vals AI

@ValsAI

Muse Spark 1.2 just cracked the top 5 on the Vals Index, at just $0.69 per test. This is 3x cheaper than Kimi and 10x or more cheaper than Fable, Opus, and 5.6 Sol.

A few things right off the bat: • it reasons very efficiently. Like Grok 4.5 levels of efficient thinking • it quantizes very well. My first few tests with iq3_xxs were better than Qwen/Gemma behaved at that size • its knowledge depth is amazing. It beats Qwen3.6 27B on no-tools trivia. • in OpenCode it is a much more efficient agent than 27B. Both models accomplish their tasks but Muse-Glimmer got there faster every time I'll say that it's worse at most things coding, probably being closer to Gemma4-31B level.. but damn there's a lot of places where I'd use this model on a 24GB GPU right now and it's been a while since anything has filled that spot except for 3.6-27B

Tested Muse Spark 1.3 on a ~10k LOC codebase and found it sometimes beat Gemini 3.8 Flash on hard multi-part backend bugs, though Gemini produced better-looking frontends.

Dilmer

@Dilmerv

Hello everyone! 👋🕶️ I put Meta Muse Code + the Unity CLI to the test with a real Unity project to see how far I could push this agentic workflow. Instead of just generating code, I wanted to see how it could actually interact with Unity, run tests, validate changes, port the h

We've been curious how far local models have actually come for agentic coding tasks, so we ran an experiment. Setup: • Model: Muse Glimmer (30B), packaged as a single llamafile • Agent: Hermes coding agent (connected via llamafile's local server mode, zero API keys needed) • Target: Mozilla AI's Otari gateway The Issue: We pointed Hermes at a real, reported bug in Otari (#183) where the gateway returned a vague 502 error on image requests instead of passing through the actual provider error. What the Agent Did: Hermes read the issue, navigated the repo, isolated the bug, created a branch, ran existing tests, wrote a new regression test, and opened a draft PR (#727). All of it ran locally and offline, with zero code written by hand. It's still draft PR territory rather than a merged fix, but it's a solid signal that ~30B local models are getting genuinely capable for real dev workflows, not just toy demos. Video walkthrough of the run: https://youtu.be/5GAgbT-XgHU?si=vJqEDGm9hssCO5-M Happy to answer questions about the setup, model performance, or how Hermes handled tool calling!

Dilmer Valecillos uses Muse Code with Muse Spark 1.2 and the Unity CLI to run tests, validate changes, port a Mini Golf game to other platforms and convert it to VR.

Bijan Bowen runs Muse Spark 1.3 through a browser OS, C++ skate game, subway FPS, 3D CAD Hot Wheels, a Blender/Godot wrestling game and more, then tallies the total cost.

🚨 AI News | TestingCatalog

@testingcatalog

Since the Meta Muse agent isn't available to everyone, here is a quick UI walkthrough of its web version. Highlights 👀 > The animated Muse avatar is very cool! You can ask Muse to change how it looks, and it will generate the new look along with all the animations. >

WorldofAI benchmarks Muse Glimmer 30B on consumer hardware with a self-built test harness and compares it against Qwen 3.6 27B.

AI Coding Daily re-runs its Muse Spark 1.3 tests using the max reasoning level inside Muse Code.

hAcx

@hacx

I need to talk about how insane Muse is for a sec My friend was stressing about the Life in the UK test — the one you need for indefinite leave to remain. It's just tons of random facts about British history, government, traditions. He was drowning in the handbook. So I told

Deedy

@deedydas

The coolest thing Meta AI's Muse Spark can do by far is counting objects! As you can tell, it's far from perfect. They call it "visual grounding" and it can count objects and do bounding boxes. I've been playing with the new model and here's what I think so far: Good stuff: –

Evergreen Capital

@evergreencap3

I tested $META's Muse Spark over the last few hours and came away net positive. 3 main takeaways: 1) Quality: It's a very good model. Not quite frontier but good. It showed comparable performance vs Opus 4.6 across web data search, PDF parsing, and general

Dilmer

@Dilmerv

Ok, here’s the VR port of my Miniature Golf prototype! ⛳🥽 This project is now running on standalone (macOS/Linux/Windows), Three.js, and as of today, VR. Crazy how fast we can move today. The VR version was fully ported using Meta Muse Code and the Unity CLI. I also now have

Fahd Mirza puts Muse Spark 1.3 through agentic and coding tasks on launch day.

Fahd Mirza installs Muse Glimmer in GGUF format and tests it with DFlash speculative decoding for faster local inference.

ROCmFP4 and ROCmFP8 builds of Muse Glimmer 30B and its drafter, targeted and tested on AMD Strix Halo (gfx1151).

MindStudio notes Muse Spark 1.3 topped DeepSWE at 75.4 and placed third on Artificial Analysis, yet in a hands-on game-clone test produced "a cube shooting at other cubes."

Four spawn experiments against Muse's multi-agent runtime produced six database lock timeouts, recovered from its durable trace.

@vcruz305

@vcruz305

Tested SGLang and vLLM launch recipes for an NVFP4 export of a Hermes-agentic fine-tune of Muse-Glimmer-30B on NVIDIA DGX Spark (GB10).

KBrew

@kbrewFL

This is very interesting. I asked Muse to add an interactive map to the page and not only did it do it, but it even wrote and executed its own test script!

@Joji228

@Joji228

A single-file Three.js superhero platformer designed, coded and play-tested by Muse Spark 1.3 at xhigh effort in one autonomous session, passing 25/25 automated browser checks.

AICodeKing tests Muse Spark 1.3 and Gemini 3.8 Flash on frontend, 3D, SVG, maths and agentic coding, and flags file-overwriting issues.

AI Coding Daily runs Meta's new Muse Spark 1.3 through a set of coding experiments to see whether it belongs among frontier coding models.

@HarjjotSinghh

@HarjjotSinghh

Named profiles for the Muse Code CLI: keep several logins on one machine, run two at once and let each project pick its profile. Published on npm and tested against Muse Code 1.3.0.

Bijan Bowen puts Muse Glimmer through a battery of builds including a browser OS, C++ skate game, subway FPS, 3D CAD model, drum-kit sim and multimodal coding.

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.

Cloud Codes tests whether Muse Glimmer on a single 24GB VRAM GPU can compete with cloud coding agents in multi-step tool-calling loops.

khaled

@eltokh7

Muse Spark 1.1 seems like a very good model. I tested it on Stata Benchmark and it ranks 4th (!) beating Opus 4.7/4.8, tied with GPT 5.5. I like this benchmark because it often catches seemingly good model performing poorly on OOD tasks

@lobanov

@lobanov

An effort to make Muse Glimmer 30B actually use a 512k-token context (4x native) as a ~17GB GGUF in 32GB VRAM, trained on DGX Spark and evaluated with RULER-style retrieval tests.

Ira Bodnar shows how to give Muse live Meta Ads campaign, performance and audience data through Ryze AI in about two minutes, with test prompts and troubleshooting.

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

@murpheycandler

@murpheycandler

A small Inspect evaluation on Muse Glimmer that tests whether incentive framing changes what an agent reports to its principal when the evidence is held constant; the author reports a null result.

Matt Johnston runs Muse Spark 1.3 blind through his benchmark: a Halo build looked frontier, but XCOM, Diablo and the multi-turn agentic test broke.

WEBdoze pairs Muse Spark 1.3 and Gemini 3.8 Flash with the Impeccable design skill and browser testing to turn plain AI-generated pages into polished frontends.

Dilmer

@Dilmerv

Last night, I built a VR game prototype in a few hours with Muse Code on PC. It was already available on macOS and Linux, and today we’re bringing it to PC! 🔥 I tested it with Unity through the Unity CLI, and here are a few things that worked really well for me: - Having Muse

Shifros

@Studioshifros

Decided to stress test Muse Spark 1.3 xhigh on 3D game logic in Three.js. Built out an entire browser driving demo with different daytimes, and vehicle physics. TBH the result is so much better for the time I spent on it. @threejs @alexandr_wang Try: gg-shifro.vercel.app

@ND-DAC-DOME

@ND-DAC-DOME

A test bed comparing Muse-Glimmer-30B against Qwen3.6-27B and Qwen3.8-27B under identical settings; with 32k-token budgets the three were about even (MMLU-Pro 82/82/80%).

Luke's Dev Lab tests Muse Glimmer on a single-GPU 16GB setup across performance, memory, agency, HumanEval, and builds like a Kanban app, sand physics, a dungeon crawler, Blender and Godot.

@krtarunsingh

@krtarunsingh

An experiment running Muse Glimmer 30B Q4_K_M via llama.cpp on an RTX 4060 laptop with 8 GB VRAM, testing autonomous Python bug fixing, tool-failure recovery and multimodal invoice extraction.

Ran the model with quants (Q4) by Unsloth with latest (build from master) llama.cpp server. It takes ~20GB ram running on M5 Pro with 48GB at about 17t/s. Didn't do any reasoning loops/overthinking. Overall, sits below Qwen3.6 27B, wasn't able to get good code (frontend and backend) results. On the positive side, it didn't fail any tool calls. Your opinions/findings? Watch more: https://www.youtube.com/watch?v=_5wKhkUT438

Tornado guy

@fanofaliens

Muse Spark 1.3 Max is out and I already love how good it is at 3D stuff. We’re so back.