Model
Muse Glimmer 30B
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.
5 providersreleased knowledge cutoff 2026-01-04open weights
$0.3
best input /M · OpenRouter
$1.10
best output /M
131K
context window
131K
max output
Compare providers (sorted by listed input price)
| Provider | Input /M | Output /M | Cache read /M | Cache write /M | Context | Capabilities | Status | Updated |
|---|---|---|---|---|---|---|---|---|
| Nvidia meta/muse-glimmer-30b | — | — | — | — | 131K | reasoningtoolsstructuredvision | Aug 10, 2026 | |
| OpenRouter meta/muse-glimmer-30b | $0.3 | $1.10 | $0.04 | — | 131K | reasoningtoolsstructuredvision | Aug 10, 2026 | |
| Kilo Gateway meta/muse-glimmer-30b | $0.3 | $1.10 | $0.04 | — | 131K | reasoningtoolsstructuredvision | Aug 10, 2026 | |
| NanoGPT meta/muse-glimmer-30b | $0.35 | $1.50 | $0.04 | — | 131K | reasoningtoolsstructuredvision | Aug 10, 2026 | |
| Vercel AI Gateway meta/muse-glimmer-30b | $0.35 | $1.50 | $0.04 | — | 131K | reasoningtoolsstructuredvision | Aug 10, 2026 |
Price badge
https://models.sutraworks.ai/badge/meta/muse-glimmer-30b.svgLive SVG, regenerated on every hourly sync — paste it into a README to always show the current cheapest listed price. Updates when prices change; no build step needed on your side.
Benchmarks
- MCP Atlas(success rate)report ↗75.5
- DeepSearch QAreport ↗74.6
- SWE-Bench Pro ↗(resolve rate)report ↗51.2
- SWE-Bench Verified ↗(resolve rate)report ↗76
- Terminal-Bench ↗(success rate)report ↗51.7
- OSWorld-Verified ↗(success rate)report ↗65.9
- AIME 2026(accuracy)report ↗94.7
- GPQA Diamond ↗(accuracy)report ↗83.5
- CharXiv Reasoning(accuracy)report ↗78.8
Weights
Prices are per million tokens (USD). “—” means the provider does not publicly list a price for this model. Data from models.dev; verify with the provider before purchasing.