Muse for local & open models.

SGLang@sgl_project · 10 Aug

SGLang is honored to provide day-0 support for @AIatMeta's Muse Glimmer. ~230 tok/s on a single RTX 5090 with NVFP4 + DFlash on, and it runs out of the box on @NVIDIAAI RTX PRO 6000, DGX Spark, and Apple Silicon via MLX. Huge thanks to the NVIDIA and Meta teams for the

Xuan-Son Nguyen@ngxson · 10 Aug

We are happy to announce that Muse Glimmer is day-0 supported on llama.cpp. Meta also provides an official GGUF quant: https://t.co/7uDSddzqKZ

TimDarcet@TimDarcet · 10 Aug

Happy to release ✨ Muse Glimmer ✨ - level ~= Qwen 3.6-27B - Apache 2 - 30B dense - quantized to run in 17GB - quant + spec dec => 50 tok/s on macbook m5 max, interactive, smooth Enjoy! https://t.co/5Y8FzeY0R9

atomic.chat@atomic_chat_hq · 11 Aug

Run Meta's new Muse Glimmer 30B♾locally with 16GB VRAM! We ship our own GGUF quants. AD-IQ3_XXS does 62 tokens/s on a single RTX 4080 with vision and DFlash, and picks the same next token as the BF16 original 90% of the time! Run the model via Atomic Chat https://t.co/3IcSsjB48W

Paolo Rosson@redp314 · 10 Aug

Got Meta's new Muse Glimmer 30B running on my MacBook (M3 Max, 96GG) and tested the serving options available so far. Fastest right now: Ollama's MLX engine (DFlash included) at ~29 tok/s. Tuned llama.cpp: ~21. Raw mlx-vlm: ~10, not optimized yet. Numbers below if you're https://t.co/bvMgkQhk8L

Together AI@togethercompute · 10 Aug

Muse Glimmer is now live on Together AI. We’re proud to be a Day 0 launch partner for this open-weight model from Meta Superintelligence Labs, built for long-running agents that can reason, use tools, recover, and keep working across complex tasks. https://t.co/QP8KvRgINN

kwindla@kwindla · 11 Aug

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 https://t.co/2EmgmCzfVY

Sumanth@Sumanth_077 · 12 Aug

Run and fine-tune Meta's Muse Glimmer locally! Meta released Muse Glimmer, a 30B dense vision model designed for local agentic and coding workflows. The first open model from Meta Superintelligence Labs, released under Apache 2.0. The model runs locally at different memory https://t.co/rFaKMaikar

filipe@filicroval · 10 Aug

got Muse Glimmer 30B running on a single DGX Spark / GX10 at 29.6 tok/s with DFlash, up from 11.9 tok/s baseline, 2.49× raw decode throughput. packaged the whole thing into a reproducible GB10 recipe: - pinned weights + SHA-256s - llama.cpp compiled for GB10 / SM121 - baseline https://t.co/0Updx3waaN

Ryan | Intelligence Engineer@RyanIntEngineer · 16 Aug

I just built and validated a complete Muse Glimmer deployment for one DGX Spark. I didn’t make another quant—I integrated the available pieces into a pinned, verified, DFlash-accelerated, tool-capable, reboot-durable stack that actually works in OWUI.

davidlee@davidlee · 10 Aug

Muse Glimmer powering any harness including Codex and Claude Code using @ollama

松xR@matsu_vr · 10 Aug

muse glimmer 30Bを公式の17GBに収まるよバージョンのggufでLM Studioで動かしてみました。写真の批評もバッチリしてくれる。これは僕が撮った写真なのでネットにないのでちゃんと写真を見ているはず。コンテクスト長64000にして、VRAM20GB以内に収まっているので、かなり実用的かもしれない！ https://t.co/mV2qNCrSju

Raj@hrhraj · 18 Sept

@dps @Muse Took your custom-connector idea to the physical world: the "service with an API" was my robot. Muse wrote the connector for my rover's API, then installed PyTorch & Depth Anything V2 in the VM because the camera is 2D, drove to the black ball and stopped a few inches short. https://t.co/AcXHSxDcic

▶