Muse Glimmer 30B
muse
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.
ReasoningTool CallingAttachmentsStructured OutputOpen WeightsApache 2.0
Context Window
131K
Max Output
131K
Temperature
Yes
Open Weights
Yes
Knowledge Cutoff
2026-01-04
Released
2026-08-10
Last Updated
2026-08-10
License
Apache 2.0
Modalities
Input:TextImage
→Output:Text
Benchmarks
| Benchmark | Score | Metric |
|---|---|---|
| MCP Atlas | 75.5 | success rate |
| DeepSearch QA | 74.6 | |
| SWE-Bench Pro | 51.2 | resolve rate |
| SWE-Bench Verified | 76 | resolve rate |
| Terminal-Bench | 51.7 | success rate |
| OSWorld-Verified | 65.9 | success rate |
| AIME 2026 | 94.7 | accuracy |
| GPQA Diamond | 83.5 | accuracy |
| CharXiv Reasoning | 78.8 | accuracy |
Links
Available Providers (5)
| Provider | Input /1M | Output /1M | Cache Read /1M | Cache Write /1M | Reasoning | Status |
|---|---|---|---|---|---|---|
 NanoGPT | Rp 6.259 | Rp 26.821 | Rp 716 | — | effort | — |
 Kilo Gateway | Rp 5.365 | Rp 19.669 | Rp 716 | — | effort | — |
 OpenRouter | Rp 5.365 | Rp 21.457 | Rp 716 | — | effort | — |
 Vercel AI Gateway | Rp 6.259 | Rp 26.821 | Rp 716 | — | effort | — |
 Nvidia | Gratis | Gratis | — | — | effort | — |