Muse Glimmer 30B

Tool callingStructured outputReasoningImage input

Muse Glimmer is a 30-billion-parameter open-weight multimodal model from Meta Superintelligence Labs, distilled from Muse Spark for always-on local agents, tool use, coding, and image understanding.

Context: 131K
Released: 2mo ago
Parameters: 30B
Input / 1M tokens: $0.30
Output / 1M tokens: $1.20
Cached input / 1M tokens: $0.04

Sources:models.devOpenRouter

Benchmarks

DeepSWE is the primary coding benchmark where an exact model evaluation exists. Other sourced evaluations appear below; missing scores are not zero.

Not evaluated in DeepSWE 1.1 as of 2026-09-22. Results are only shown for exact model versions in the official dataset. Official leaderboard.

No benchmark results have been matched to this model yet.

Specifications

Developer: Meta
Model family: Muse
Type: Language model
Predecessor: —
Successor: —
Released: August 10, 2026
Model ID: meta/muse-glimmer-30b
Openness: Open weights
License: Apache 2.0
Parameters: 30B
Input modalities: text, image
Output modalities: text
Modality: text+image->text
Native context: 128K
Tokenizer: Other
Max output tokens: 16K
Knowledge cutoff: 2026-01-04
Hugging Face: meta-models/Muse-Glimmer-30B
Source freshness: Metadata synced Oct 8, 2026, 10:17 PM UTC

Open-weight builds

Ollama variants

Every published Ollama tag matched to this model, including parameter editions, instruction/base variants, and quantized builds.

12 builds

Variant	Parameters	Type	Quantization	Size	Context	Input
muse-glimmer:30b-bf16	30B	—	BF16	57 GB	128K	text + image
muse-glimmer:30b-bf16-dflash	30B	—	BF16	59 GB	128K	text + image
muse-glimmer:30b-mlx-bf16	30B	—	BF16	60 GB	128K	text + image
muse-glimmer:30b-mlx-bf16-dflash	30B	—	BF16	65 GB	128K	text + image
muse-glimmer:30b-q4_K_M-dflash	30B	—	Q4_K_M	20 GB	128K	text + image
muse-glimmer:30b-q8_0	30B	—	Q8_0	31 GB	128K	text + image
muse-glimmer:30b-q8_0-dflash	30B	—	Q8_0	33 GB	128K	text + image
muse-glimmer:30b-mlx	30B	—	—	19 GB	128K	text + image
muse-glimmer:30b-mxfp8	30B	—	—	33 GB	128K	text + image
muse-glimmer:30b-mxfp8-dflash	30B	—	—	35 GB	128K	text + image
muse-glimmer:30b-nvfp4	30B	—	—	17 GB	128K	text + image
muse-glimmer:latest	—	—	—	18 GB	128K	text + image

Additional source evidence

Source model: meta/muse-glimmer-30b
Source family: muse
Last updated: 2026-08-10
Knowledge cutoff: 2026-01-04
Model context: 131K
Model output limit: 131K
Open weights: Yes
File attachments: Yes
Reasoning: Yes
Tool calling: Yes
Structured output: Yes
Temperature control: Yes

Downloadable weights

Hugging Face

Published benchmarks

Benchmark	Score	Evaluation details	Evidence
MCP Atlas	75.5 · success rate	public · 2026-08-10	Source
DeepSearch QA	74.6	2026-08-10	Source
SWE-Bench Pro	51.2 · resolve rate	2026-08-10	Source
SWE-Bench Verified	76 · resolve rate	2026-08-10	Source
Terminal-Bench	51.7 · success rate	with terminus2 · 2.1 · 2026-08-10	Source
OSWorld-Verified	65.9 · success rate	2026-08-10	Source
AIME 2026	94.7 · accuracy	2026-08-10	Source
GPQA Diamond	83.5 · accuracy	AA · 2026-08-10	Source
CharXiv Reasoning	78.8 · accuracy	2026-08-10	Source

Official references

Complete source record (JSON)

{
  "id": "meta/muse-glimmer-30b",
  "name": "Muse Glimmer 30B",
  "description": "Muse Glimmer is a 30-billion-parameter open-weight multimodal model from Meta Superintelligence Labs, distilled from Muse Spark for always-on local agents, tool use, coding, and image understanding.",
  "family": "muse",
  "last_updated": "2026-08-10",
  "knowledge": "2026-01-04",
  "attachment": true,
  "reasoning": true,
  "tool_call": true,
  "structured_output": true,
  "temperature": true,
  "modalities": {
    "input": [
      "text",
      "image"
    ],
    "output": [
      "text"
    ]
  },
  "limit": {
    "context": 131072,
    "output": 131072
  },
  "benchmarks": [
    {
      "name": "MCP Atlas",
      "score": 75.5,
      "source": "https://huggingface.co/meta-models/Muse-Glimmer-30B",
      "metric": "success rate",
      "variant": "public",
      "date": "2026-08-10"
    },
    {
      "name": "DeepSearch QA",
      "score": 74.6,
      "source": "https://huggingface.co/meta-models/Muse-Glimmer-30B",
      "date": "2026-08-10"
    },
    {
      "name": "SWE-Bench Pro",
      "score": 51.2,
      "source": "https://huggingface.co/meta-models/Muse-Glimmer-30B",
      "metric": "resolve rate",
      "date": "2026-08-10"
    },
    {
      "name": "SWE-Bench Verified",
      "score": 76,
      "source": "https://huggingface.co/meta-models/Muse-Glimmer-30B",
      "metric": "resolve rate",
      "date": "2026-08-10"
    },
    {
      "name": "Terminal-Bench",
      "score": 51.7,
      "source": "https://huggingface.co/meta-models/Muse-Glimmer-30B",
      "metric": "success rate",
      "variant": "with terminus2",
      "version": "2.1",
      "date": "2026-08-10"
    },
    {
      "name": "OSWorld-Verified",
      "score": 65.9,
      "source": "https://huggingface.co/meta-models/Muse-Glimmer-30B",
      "metric": "success rate",
      "date": "2026-08-10"
    },
    {
      "name": "AIME 2026",
      "score": 94.7,
      "source": "https://huggingface.co/meta-models/Muse-Glimmer-30B",
      "metric": "accuracy",
      "date": "2026-08-10"
    },
    {
      "name": "GPQA Diamond",
      "score": 83.5,
      "source": "https://huggingface.co/meta-models/Muse-Glimmer-30B",
      "metric": "accuracy",
      "variant": "AA",
      "date": "2026-08-10"
    },
    {
      "name": "CharXiv Reasoning",
      "score": 78.8,
      "source": "https://huggingface.co/meta-models/Muse-Glimmer-30B",
      "metric": "accuracy",
      "date": "2026-08-10"
    }
  ],
  "release_date": "2026-08-10",
  "open_weights": true,
  "license": "Apache 2.0",
  "links": [
    {
      "url": "https://research.meta.ai/blog/introducing-muse-glimmer-open-agentic-model",
      "label": "Announcement",
      "type": "announcement"
    },
    {
      "url": "https://huggingface.co/meta-models/Muse-Glimmer-30B",
      "label": "Model card",
      "type": "model_card"
    },
    {
      "url": "https://developer.meta.com/ai/models/muse-glimmer/",
      "label": "Developer docs",
      "type": "docs"
    }
  ],
  "weights": [
    {
      "label": "Hugging Face",
      "url": "https://huggingface.co/meta-models/Muse-Glimmer-30B"
    }
  ]
}

Infrastructure

Provider availability

All providers

Provider	Endpoint	Context	Input /M tokens	Output /M tokens	Cached input /M tokens	Cache write /M tokens	Uptime 1d	Data handling
DeepInfra	deepinfra/bf16	131K	$0.30	$1.20	$0.04	—	100.00%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
Together	together	131K	$0.35	$1.50	$0.04	—	80.46%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15

Provider data:OpenRouter

Deployment graph

Can I run this?

Vertical Tabs only shows compatibility when it has structured evidence tying this exact model to hardware, runtime, quantization, memory use, and observed performance.

Browse hardware