The Open Weights
LatestModelsLeaderboardsCompanies
Subscribe
The Open Weights

The daily record of open-source AI. New model releases, leaderboards, and what's coming next — written for people who ship.

Refreshed every 12 hours

Discover

  • Latest releases
  • New today
  • Trending models

Browse

  • All models
  • Companies
  • Categories
  • Leaderboards

About

  • About
  • Editorial policy
  • Hardware estimates
  • RSS feed
  • llms.txt
  • Newsletter

© 2026 The Open Weights. An independent publication.

PrivacyTermsSMSAggregated by Claude · curated by humans.
LatestMeta AIGlimmer 30B
Meta AIVision-Language

Meta's Muse Glimmer 30B Targets Local Agentic Coding

A 30-billion-parameter multimodal model built to run locally, released under Apache 2.0 with an eye on agentic coding workflows.

Aug 10, 2026
Major releaseApache 2.0
Muse Glimmer 30B

Meta AI has released Muse Glimmer 30B, a 30-billion-parameter model that blends vision, code, and reasoning capabilities in a single dense architecture. The company is positioning it as an open, local-first tool for agentic coding — the kind of workflow where a model plans, reads context, and executes tasks with limited hand-holding.

Unlike a mixture-of-experts design, Glimmer is a dense 30B model, which keeps its behavior predictable and its footprint manageable for teams that want to run it on their own hardware rather than through an API. It ships under the permissive Apache 2.0 license, allowing commercial use and modification without the usage caveats attached to some other open-weight releases.

Why it matters

The interesting move here is the combination of traits Meta is emphasizing:

  • Multimodal (VLM): it can process images alongside text, useful for reading screenshots, diagrams, or UI states in a coding loop.
  • Agentic: framed for multi-step, tool-using workflows rather than one-shot chat.
  • Local and open: the 30B size and Apache license make on-premises deployment realistic for privacy-conscious or cost-sensitive teams.

That mix puts Glimmer in direct conversation with the growing class of open coding assistants, where the pitch is less about raw leaderboard supremacy and more about control — running capable models close to your data without vendor lock-in.

Meta has not published a context length or detailed specifications in the record accompanying this release, and independent benchmarks will ultimately determine how Glimmer stacks up against comparable open models. For now, the details and weights are available through Meta's announcement on Hugging Face.

Sources

  • Meta is back with Muse Glimmer: local, agentic, multimodal, and open source

    Announcement

    Visit
  • meta-models/Muse-Glimmer-30B

    Hugging Face

    Visit
  • Meta releases Muse Glimmer, 30B parameter open-weight edge model

    Hacker News

    Visit
Olderdots3-note preview brings audio and vision to one modelDots Studio · Any-to-Any · 3 months agoNewerinclusionAI Releases Ling-3.0-tiny, an MIT-Licensed MoEinclusionAI · Text / LLM · 3 months ago

Get the model

Hugging FaceAnnouncementHacker News

Specs

Parameters30B
Context window131K tokens
Size59.6 GB
PrecisionBF16
ArchitectureMuseGlimmerForConditionalGeneration
LicenseAPACHE-2.0
Downloads150.2K
Likes2.2K

Can you run it?

Runs on a 24 GB GPU at 4-bit.

  • BF16 (as published)

    80 GB GPU (A100 / H100) · 128 GB Mac Studio

    61 GB
  • 8-bit

    48 GB GPU (RTX 6000) or 2×24 GB · 64 GB Mac

    33 GB
  • 4-bit

    24 GB GPU (RTX 3090 / 4090) · 32 GB Mac

    19.3 GB
Your machine
GGUF builds
  • KV cache sized for 8,192 tokens of context; longer prompts need proportionally more.

Based on Hugging Face weights metadata. Assumes an 8K context and ~1 GB runtime overhead; actual needs vary. How we estimate


Modalities

Vision-LanguageCodeReasoning

The Weekly Weights

Every open release that mattered, one email a week.

0 comments

No comments yet. Be the first to weigh in.

More from Meta AI

All Meta AI releases →
Meta AI/Text / LLM

Meta AI debuts Muse Ember, a compact open model

The company's smallest, most efficient open-weight language model targets lean, on-device text workloads.

Nov 10, 2026
Meta AI/Code

Meta's Muse Spark Open Targets Agentic Coding

The open-weight model pairs a million-token context window with a focus on long-horizon code tasks.

Oct 15, 2026
Meta AI/Any-to-Any

Meta's Muse Spark Open Targets Agentic Coding

The new open-weight model pairs a million-token context window with Meta's bet on long-horizon coding agents.

Oct 15, 2026

More in Vision-Language

All Vision-Language →
SenseNova-Loop-9B
SenseTime/Vision-LanguageRuns on a laptop

SenseTime Releases Looped 9B Multimodal Model

The company's new SenseNova-Loop-9B pairs vision-language capabilities with a recurrent looped-transformer design.

Nov 17, 2026
Hunyuan Hy4-flash
Tencent/Vision-LanguageMulti-GPU server

Tencent releases Hunyuan Hy4-flash multimodal MoE

The lightweight, mixture-of-experts model adds vision and text instruction-following to Tencent's open Hunyuan lineup.

Nov 17, 2026
Mirage
Cactus Compute/Vision-Language

Cactus Compute's Mirage brings vision tasks on-device

The new open vision-language model targets detection, segmentation, and visual Q&A without relying on the cloud.

Nov 5, 2026