Always-on agents
Built for agents that don’t stop with reliable tool-calling, persistent state across restarts, and self-managed memory across hours-long sessions.
Muse Glimmer · Open model for always-on local agents 50 Starter Credits
Muse Glimmer
An open model built for always-on local agents. 30B parameters, licensed under Apache 2.0 and runs on a single GPU — tuned for tool use, long tasks, and failure recovery.
30B parametersRuns on a single GPU or MacTool use · Long tasks · Failure recoveryMultimodal perception built in
Sign in with Google to start. Agentic sessions run on credits.
Muse Glimmer · Agentic session
Investigate the checkout test flakes in my repo and open a PR with the fix.
Reads context and prior results in the worktree…
Runs the failing suite in an isolated worktree — 3 flakes reproduced…
Opens PR #284 — 3 files changed, tests green ✓
Tool callingPersistent stateHours-long sessions30B · Apache 2.0
Meet Muse Glimmer
Built to keep working, built to run anywhere
Muse Glimmer is trained to deliver competitive agentic and coding performance, with multimodal perception built in.
Optimized for local deployments
It’s small enough to run on a single consumer GPU or Mac, enabling use cases that range from local agents to local coding.
Multimodal perception
Muse Glimmer is trained to deliver competitive agentic and coding performance, with multimodal perception built in.
What it looks like
One session, from first task to finished result
Start an agentic session in your browser — paste a prompt, add context, and let Muse Glimmer build.
Local coding
Small enough to run on a single consumer GPU or Mac, enabling use cases that range from local agents to local coding.
Tool use that recovers
Reliable tool-calling, persistent state across restarts, and self-managed memory across hours-long sessions.
Agentic + coding performance
Competitive agentic and coding performance with multimodal perception built in — from planning to pull request.
How to use Muse Glimmer
Start an agent session in four steps
Create an account, describe a task, and let the model plan, call tools, and keep working.
Sign in with Google
Create your account and your 50 starter credits are ready to use.
Pick a task
Describe an agentic job — a coding change, a research run, or a long-horizon workflow.
Muse Glimmer plans and executes
The model calls tools, tracks context and prior results, and keeps working across hours-long sessions.
Review, continue, or top up
Check the results in your session, keep the agent going, or add credits when you need them.
Benchmarks
Muse Glimmer at a glance
Muse Glimmer is trained to deliver competitive agentic and coding performance, with multimodal perception built in.
| Benchmark | Muse Glimmer-30B High reasoning | Gemma4-31B Thinking mode | Qwen3.6-27B Thinking mode |
|---|---|---|---|
| General agentic | |||
| MCP Atlas | 75.5 | 54.2 | 62.5 |
| DeepSearch QA | 74.6 | 61.7 | 71.1 |
| τ3-Banking | 23.5 | 15.1 | 16.7 |
| WildClawBench | 47.6 | 37.6 | 43.2 |
| GDPval-AA | 953 | 811 | 1141 |
| GAIA2 | 43.3 | 36.4 | 40.0 |
| SkillsBenchWith skills | 44.3 | 32.4 | 46.6 |
| OSWorld-Verified | 65.9 | 58.5 | 75.6 |
| Agentic coding | |||
| SWE-Bench Pro | 51.2 | 36.9 | 50.2 |
| SWE-Bench Verified | 76.0 | 66.6 | 77.2 |
| TerminalBench 2.1 | 51.7 | 43.4 | 60.7 |
| SciCode | 43.6 | 43.4 | 39.8 |
| Multimodal | |||
| Charxiv Reasoning | 78.8 | 77.7 | 78.4 |
| ScreenSpot Pro | 75.4 | 75.9 | 76.1 |
| OmniDocBench v1.5 | 75.8 | 72.5 | 77.8 |
| MMMU Pro | 74 | 73 | 75 |
| Safety | |||
| CI Memories | Violation ↓: 26.4 Coverage: 64.8 | Violation ↓: 12.1 Coverage: 53.0 | Violation ↓: 53.4 Coverage: 66.9 |
| Siren AgentDojo | Attack Success Rate ↓: 28.4 Utility: 94.2 | Attack Success Rate ↓: 25.6 Utility: 90.8 | Attack Success Rate ↓: 40.3 Utility: 92.7 |
| General capabilities and reasoning | |||
| IFBench | 77.0 | 76.0 | 70.8 |
| AIME 2026 | 94.7 | 89.2 | 94.1 |
| GPQA Diamond | 83.5 | 85.7 | 84.2 |
| Humanity's Last ExamText, No tools | 22.0 | 23.6 | 23.1 |
| AA-LCR | 80.0 | 68.3 | 73.3 |
| Beam 128K | 65.1 | 58.2 | 63.0 |
For the full model card, please see the Muse Glimmer page on Hugging Face.
For more details about Muse Glimmer evaluations, read the methodology report.
Meta is committed to promoting safe and fair use of its tools and features, including Muse Glimmer.
Where to run Muse Glimmer
Run it on the platform you already use
Deploy Muse Glimmer with transformers, llama.cpp, vLLM, Colab, or any of these popular providers.
Hugging Face
Run Muse Glimmer with transformers, llama.cpp, vLLM, Colab, and other popular platforms.
Ollama
Run Muse Glimmer on your own hardware in one simple, seamless CLI command.
Unsloth
Run, fine-tune, and deploy Muse Glimmer from one unified local UI interface with the open-source Unsloth Desktop app.
LM Studio
LM Studio Bionic is the easiest way to run Muse Glimmer for agentic tasks, locally on your computer.
Fireworks
Fireworks offers production-grade Muse Glimmer deployments with precise reasoning controls and effortless scaling for unpredictable agent traffic.
together.AI
Run Muse Glimmer on Together Serverless Inference for long-running agentic workflows.
OpenRouter
Run Muse Glimmer via a single unified API that routes across multiple inference providers for cost, speed, and availability.
Prefer the hosted experience?
Start an agentic session in your browser on museglimmer.lol.
Plans and credits
Choose a Muse Glimmer Credit Pack
Buy credits once — purchased credits never expire, and agent sessions run on credits.
Starter
$9.90one time
Get started with a one-time credit pack.
370credits
- One-time purchase, no subscription
- Purchased credits never expire
- Runs on Muse Glimmer agent sessions
- Sign in with Google
Creator
$29.90one time
The popular choice for regular agent sessions.
1,300credits
- One-time purchase, no subscription
- Purchased credits never expire
- Runs on Muse Glimmer agent sessions
- Sign in with Google
Studio
$49.90one time
For power users running long agentic workflows.
2,500credits
- One-time purchase, no subscription
- Purchased credits never expire
- Runs on Muse Glimmer agent sessions
- Sign in with Google
Business
$99.90one time
Best value for teams and heavy usage.
5,550credits
- One-time purchase, no subscription
- Purchased credits never expire
- Runs on Muse Glimmer agent sessions
- Sign in with Google
Compare credit usage, billing details, and all available plans on the Muse Glimmer Pricing page.
Frequently asked questions
Muse Glimmer FAQ
Quick answers about Muse Glimmer, local agents, credits, and sessions.
What is Muse Glimmer?+
Muse Glimmer is an open model built for always-on local agents — 30B parameters, licensed under Apache 2.0, and tuned for tool use, long tasks, and failure recovery.
Can I run Muse Glimmer on my own machine?+
It’s small enough to run on a single consumer GPU or Mac, enabling use cases that range from local agents to local coding.
How do sessions and credits work?+
Sign in with Google to start. Your account includes 50 starter credits, and agentic sessions run on credits. When your balance runs out, choose a credit pack on the Pricing page.
Is Muse Glimmer multimodal?+
Muse Glimmer is trained to deliver competitive agentic and coding performance, with multimodal perception built in.
What about long tasks and interruptions?+
Muse Glimmer is built for agents that don’t stop — with reliable tool-calling, persistent state across restarts, and self-managed memory across hours-long sessions.
What license applies to Muse Glimmer?+
Muse Glimmer is licensed under Apache 2.0. Meta is committed to promoting safe and fair use of its tools and features, and a usage policy applies to access or use of Muse Glimmer.
Start building with Muse Glimmer today.
Sign in with Google, check your credits, and jump straight into a Muse Glimmer agent session.