Muse Glimmer 30B
meta/muse-glimmer-30bMeta’s open-weight model for text, images, reasoning, and autonomous-agent tasks.
131K tokens of context — about 98,304 words, on a scale that ends at 2M.
What it can do
- Reasoning
- Vision
- Tool use
- Open weights
- Takes text
- Takes image
- Gives text
Released 9 Aug 2026 · checked 3 Oct 2026.
About it
Muse Glimmer 30B is an open-weight model from Meta Superintelligence Labs. It accepts text and images, produces text, and includes reasoning, vision, and tool-use abilities. Its 131,072-token context suits long inputs, while open weights support deployment on consumer hardware. Input costs $0.35 per million tokens; output costs $1.5 per million tokens.
Review
In a line
Muse Glimmer 30B is an open-weight multimodal model made by Meta. It works with text and images, returns text, and is designed for autonomous-agent tasks on consumer hardware.
What it's good at
The model supports reasoning and tool use, so it fits tasks where software needs to plan or act over multiple steps. Its vision ability lets it handle images as well as text.
It is intended for long-horizon work and can be deployed with its available weights rather than only accessed as a closed service.
Price in plain terms
Input costs $0.35 per million tokens. Output costs $1.5 per million tokens.
For scale, reading a 500-word page costs about 0.023 cents. A chat turn with about 1,000 input tokens and 500 output tokens costs about 0.11 cents. One thousand such turns cost about $1.10.
Context and limits
The context window is 131,072 tokens, allowing a single request to include a very large amount of input. A single response can contain up to 117,964 tokens.
It produces text only, even though it accepts both text and images.
Ways to use it
Muse Glimmer 30B is aimed at autonomous-agent software that can call tools while working through a task. Its open weights also allow deployment on consumer hardware.
The model identifiers are meta/muse-glimmer-30b and meta-models/Muse-Glimmer-30B.
Verdict
Choose this model if you need image-and-text input, reasoning, tool use, and a very large context window, or if open weights and consumer-hardware deployment matter. Its $0.35 input price also suits applications that read many tokens.
Avoid it if your application needs image or audio output, because the model returns text only. The facts also do not list specific apps, chat tools, or providers, so check that your intended software supports it.
Verdict
Muse Glimmer 30B fits builders and software users who want one model to read text and images, reason through extended tasks, and use tools. Its open weights may suit deployments on consumer hardware, while the large context window handles long inputs. A typical 1,000-token-in, 500-token-out turn costs about 0.11 cents. It is less suitable when you need generated images or other non-text output, or when your chosen app has not confirmed support for this model.
- Accepts both text and image inputs
- Supports reasoning and tool use
- 131,072-token context window
- Open weights for consumer-hardware deployment
- About $1.10 for 1,000 typical chat turns
- Produces text output only
- Maximum response length is 117,964 tokens
- No specific app or provider support is listed
- Pick it ifAutonomous-agent tasks using text and images; Long inputs and open-weight deployment; Low-cost, repeated chat workloads
- Skip it ifYou need image, audio, or other non-text output; Your software does not support this model
Pricing
Input is $0.35 per million tokens and output is $1.5 per million tokens. A typical 1,000-token-in, 500-token-out chat turn costs about 0.11 cents; 1,000 such turns cost about $1.10.
Questions
Who made Muse Glimmer 30B?
Meta Superintelligence Labs made it; its model family is Meta.
Can it understand images?
Yes. It accepts image and text inputs and has vision capability.
What does it output?
It outputs text.
How much context can it handle?
Its context window is 131,072 tokens.
How much can one answer contain?
The maximum output is 117,964 tokens.
Is it open weight?
Yes. Its weights are open, and the model is intended for deployment on consumer hardware. See open-weight models.
What does a typical chat turn cost?
A turn with about 1,000 input tokens and 500 output tokens costs about 0.11 cents. One thousand such turns cost about $1.10.
On these boards
More from Meta
- Muse Spark 1.3$1.25 in
- Muse Spark 1.3 Contributor$0.10 in
- Muse Spark 1.2 Contributor$0.10 in
- Muse Spark 1.2$1.25 in
- Muse Spark 1.1$1.25 in
- Llama Guard 4 12B$0.18 in
- Llama 4 Maverick$0.19 in
- Llama 4 Scout$0.10 in
- Llama 3.3 70B Instruct$0.10 in
- Llama 3.2 1B Instruct$0.027 in
- Llama 3.2 3B Instruct$0.050 in
- Llama 3.1 70B Instruct$0.40 in
- Llama 3.1 8B Instruct$0.050 in