Local & open models

@HawgAuto

@HawgAuto

Glimmer HD Vision is an OpenAI-compatible proxy that keeps images within Muse Glimmer 30B's 4,096 visual-token limit by sending a 4K image as one overview plus four overlapping detail tiles, with an OCR/layout mode.

A 4-byte GGUF header edit that raises Muse Glimmer 30B's 131,072 context limit to 262K. The author shows YaRN scaling only adds prefill cost on this architecture, and verified 262K with 8 slots on one RTX 4090.

@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.

@mjsabby

@mjsabby

An OpenAI- and Anthropic-compatible serving stack for Muse Glimmer 30B on one or two Intel Arc Pro B70s, built oracle-first with a float64 CPU reference that gates every GPU kernel.

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.

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…

Experimental OpenVINO INT4 conversion of Muse Glimmer 30B that needs development builds of Optimum Intel and OpenVINO.

@alpharomercoma

@alpharomercoma

A reproducible setup serving Muse Glimmer 30B on one AMD MI300X with a custom 11 GB vLLM 0.28.0 ROCm image, exposed as an authenticated OpenAI-compatible endpoint and driven from the pi CLI.

turboderp's self-calibrated EXL3 quants of Muse Glimmer 30B down to 1.75 bits per weight, with calibration and eval traces.

A DPO fine-tune of Muse Glimmer focused on prose quality and humanlike characters, trained on roughly one-third writing and two-thirds roleplay data.

@Satgoy152

@Satgoy152

A project that fine-tunes a DSpark speculator for Muse Glimmer 30B on on-policy coding and agentic traces to raise acceptance length on agentic workloads, evaluated with Terminal-Bench.

@sxuff

@sxuff

A pinned llama.cpp recipe for Muse Glimmer 30B on one NVIDIA GB10 that verifies Meta's official GGUFs by SHA-256 and reproduces text, tool-call, coding, vision and throughput checks.

@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.

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.

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

Hey Guys, I've got 4 x v100's in a Dell C4140 (NVlink) and I have been working on a fork of llama.cpp that is targeted at the v100's. Looking for testers to give it a go and provide feedback. WyvernTKC/llama.cpp-4xV100: Fork of llama.cpp Nvida Volta V100 (tensor parallelism 4 x v100 GPU) model arch type size (GB) pp layer pp tensor change tg layer tg tensor change glm4 9B Q8_0 glm4 dense 9.3 1187.9 2924.3 +146 % 67.6 123.0 +82 % qwen35 27B Q8_K_P qwen35 dense 29.3 640.4 1717.6 +168 % 22.2 52.4 +137 % gemma4 31B Q8_0 gemma4 dense 30.4 679.8 1621.2 ±322 noisy 20.6 46.7 +127 % muse-glimmer 30B F16 muse-glimmer dense 51.9 1048.4 2395.6 +128 % 15.1 40.4 +168 % llama 70B Q8_0 llama dense 69.8 302.5 950.5 +214 % 9.8 28.6 +192 % qwen35moe 35B-A3B Q8_0 qwen35moe MoE 256×8 34.4 1602.6 3143.7 +96 % 93.6 113.7 +21 % qwen3next 80B-A3B Q4_K_M qwen3next MoE 512×10 45.9 889.8 1647.0 +85 % 76.6 86.8 +13 % deepseek4 284B Q2_K deepseek4 MoE 256×6 90.9 188.8 616.1 +226 % 27.3 37.7 +38 % Thanks!

@schererstefan

@schererstefan

glimmer-cli is a local TypeScript CLI for Muse Glimmer 30B and Muse Spark 1.2 via Ollama, with stubbed tools and a reproducible tool-use eval harness.

Hugging Face's launch post covers day-0 transformers, llama.cpp and vLLM support, Inference Endpoints, speculative decoding, TRL fine-tuning and agent demos for Muse Glimmer.

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.

Mark Zuckerberg

@finkd

Today we're also opening the weights for Muse Glimmer, a great 30B parameter dense model that can run locally. Soon we'll also release the weights for Muse Spark 1.2, our latest foundation model. Meta is a strong supporter of open source and I'm proud of these releases. Congrats

Meta's developer post on running Muse Glimmer on a single consumer GPU with vLLM, llama.cpp and ExecuTorch, with quantized builds in 24-32 GB and cookbook recipes.

These numbers were captured during a real feature implementation task in Next.js and Nest.js (adding a theme switching system across components). The structural predictability of UI/state refactoring is likely why DFlash hit such a high draft acceptance rate (~97%). Here is a quick log analysis and performance summary running Muse-Glimmer-30B (UD- Q6_K_XL) paired with DFlash (Speculative Decoding) via llama.cpp (llama-server + single RTX 5090). -ngl 99 -c 200000 --host 0.0.0.0 --port 8080 --timeout 600 --cache-reuse 256 --parallel 1 --flash-attn on --spec-type draft-dflash --spec-draft-n-max 16 --spec-draft-p-min 0.7 --spec-draft-ngl 99 --cache-type-k q8_0 --cache-type-v q8_0 --no-webui --load-mode none --cache-ram 12192 --temperature 0.8 --top-k 30 --top-p 0.95 --min-p 0.05 --repeat-penalty 1.1 --repeat-last-n 64 --reasoning on --chat-template-kwargs {"enable_thinking":true} Compared to Qwen 3.6 27B: No Chinese language-mixing bugs, no overthinking loops, and concise responses. Its lighter memory footprint at Q6 also freed up more VRAM/RAM for a much larger context size. Metric Measured Value Notes Generation Speed (Peak) 100 – 287 tokens/sec Average ~173 t/s across all ta

@Fangyuan025

@Fangyuan025

Chaty is a private offline desktop app built on Rust and llama.cpp that runs Muse Glimmer and other open models locally with a coding agent, RAG knowledge base, deep research and voice.

@GordonWei

@GordonWei

An OpenAI-compatible /v1/chat/completions server that runs Muse Glimmer on Apple Silicon through mlx_vlm while LM Studio's bundled MLX runtime can't yet load the architecture.

Alexandr Wang

@alexandr_wang

1/ big announcement today: we will be releasing an open weight version of muse spark 1.2 soon. we also are releasing muse glimmer, a 30B agentic model with open weights under apache 2.0. muse glimmer can run on 24GB of VRAM without losing agentic reliability. 🧵

Eleven matched on/off pairs across Gemma 4 and Qwen3.6, holding model, quant, card, corpus and concurrency fixed inside each pair. Speed: 1.65x to 2.54x, every pair. Accuracy: nothing the paired intervals could separate from ordinary run-to-run movement. Muse Glimmer is the one that lost. Meta's matching DFlash drafter made the same 7900 XTX 9% slower, keeping 24.55% of drafted tokens against roughly four in five for the Gemma and Qwen heads. Acceptance fell across the run instead of warming up. Meta's model card reports 3.1x on an RTX 5090, and there are open llama.cpp issues for DFlash on AMD and under Vulkan, so I read it as the backend rather than the model. Acceptance turned out to be a poor predictor of speed. It moved under four points across five models while the multiple nearly doubled. What tracks the multiple is how bandwidth-bound the target is: a heavier quant gains more, and the two mixture-of-experts pairs gained least. Worth knowing before you benchmark anything: -md mtp-head.gguf silently disables speculation. Use -hf REPO:QUANT -hfd REPO, then read speculative from /slots and confirm it is true. Per-pair table, intervals, acceptance counters and the raw predic

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

AWQ INT4 quant of Muse Glimmer 30B calibrated on STEM and agentic data across ten languages, 24.02 GB.

@johnhalloran321

@johnhalloran321

Scripts that run Muse-Glimmer-30B with vision and tool calling on a 32GB M2 MacBook Pro via llama.cpp and Metal, without admin or sudo access, pulling Meta's official GGUFs.

@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.

atomic.chat

@atomic_chat_hq

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

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

TimDarcet

@TimDarcet

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!

Picked up China version of the Mi50 (Radeon VII) 16GB VRAM GPU for about $135. Ran it using llama.cpp Ubuntu Vulkan prebuilt binary build: b29c606e2 (10964). Used a Power Limit or 220/190 watts on the GPUs. Dual Radeon 32GB Vram and 64GB DDR4 System Dual Radeon RX 7900 GRE and Radeon VII 32gb VRAM GGUF Models: • Qwen3.6-35B-A3B-UD-Q5_K_XL.gguf • Accio-Lab_occamy-1.0-Q5_K_S.gguf • Laguna-XS-2.1-APEX-I-Balanced.gguf • NVIDIA-Nemotron-3.5-Lightning-30B-A3B-Q5_K_M.gguf • Gemma-4-31B-it-Q6_K.gguf • Ateron_Gemma-4-MoonGem-31B-Q5_K_M.gguf • Qwen3-Coder-30B-A3B-Instruct-UD-Q6_K_XL.gguf • Qwen3-VL-30B-A3B-Thinking-UD-Q6_K_XL.gguf • Qwen3-Coder-30B-A3B-Instruct-UD-Q5_K_XL.gguf • North-Mini-Code-1.0-MXFP4_MOE.gguf • GLM-4.7-Flash-Uncen-Hrt-NEO-CODE-MAX-imat-D_AU-Q6_K.gguf • Muse-Glimmer-30B-UD-Q6_K_XL.gguf • Huihui-Qwen3.8-27B-abliterated-UD-Q6_K_XL.gguf • Qwen3.8-27B-Q6_K.gguf • Qwen3.8-27B-OBLITERATED-Q5_K_M.gguf • Medgemma-27b-it-UD-Q6_K_XL.gguf • Gemma-4-26B-A4B-it-UD-Q6_K_XL.gguf • Gemma4-26B-A4B-QAT-Uncensored-HauhauCS-Balanced-Q4_K_M.gguf • GPT-OSS-20b-abliterated.i1-Q6_K.gguf Sorted by params then size model size params pp512 tg128 qwen35moe 35B.A3B Q5_K - Medium 24.76 GiB 3

Together AI

@togethercompute

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.

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

Red Hat AI's FP4 weight and activation quantized Muse Glimmer 30B.

NVFP4 W4A4 quantized weights of Muse Glimmer 30B.

Meta's own GGUF conversions of Muse Glimmer 30B for llama.cpp.

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