Entries matching “multimodal”

Meta Developers session on what Muse Spark's act-on-perception multimodality unlocks across code, physical action and video workflows, plus Muse Voice Transcribe.

Sebastian Raschka

@rasbt

Whoa, Meta released a new open-weight LLM yesterday, something that hasn't happened since the good old Llama days. Their Meta Muse Glimmer model is a 30B multimodal reasoning model with a Gemma-like architecture design. (“Glimmer” is probably a wordplay on “Spark,” the more

Abstract Extraordinary digs into the architecture behind Meta's pitch of an autonomous, multimodal agent that runs on-device with no cloud.

@nicedreamzapp

@nicedreamzapp

An architecture port adding the muse_glimmer model class (vision tower, language model, projector and image processor) to mlx-vlm, so any Muse Glimmer checkpoint runs multimodally on Apple Silicon.

@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.

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.

Abliterated Muse Glimmer 30B GGUF quant ladder, updated with a 1.63 GB abliterated DFlash drafter and a 1.40 GB multimodal projector.

AI at Meta

@AIatMeta

Muse Spark 1.2 supports a broad range of multimodal tasks, from turning visuals into working code to translating perception into physical action. It also brings robust audio-visual understanding to enable video-heavy workflows common in real-world enterprise use. Today, we’re

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.

NVIDIA hosts a Muse Glimmer 30B endpoint on build.nvidia.com with Python (OpenAI, LangChain), JavaScript and curl examples for the ~29.6B multimodal model with 131K context.