Muse Glimmer 30B
Meta · released
Meta's open-weights model: a 30-billion-parameter text-and-image model distilled from Muse Spark and built to run tool-using agents locally on a single consumer GPU.
Best for: Running locally · Local tool-using agents · Fine-tuning
- Intelligence Index
- 17.5
- #33 of 38 · at High effort
- Price per 1M tokens
- $0.33 in $1.35 out
- Blended $0.58 at 3:1
- Context window
- 131K
- Output speed
- 89 tokens/s
- $0.06 measured cost per task
Capability profile
One 0–100 score per area, placing this model between the weakest and strongest tracked model across that area's benchmarks.
- Overall25 · #34 of 38
Based on 2 of 3 benchmarks
- Coding13 · #33 of 39
Based on 3 of 5 benchmarks
- Agents & knowledge work21 · #32 of 40
Based on 1 of 6 benchmarks
- Reasoning & science16 · #32 of 39
Based on 2 of 2 benchmarks
- Knowledge & reliability43 · #33 of 39
Based on 2 of 2 benchmarks
- Vision57 · #21 of 31
Based on 1 of 2 benchmarks
Benchmark results
All third-party results on record. Rank is among the tracked models that have been tested on that benchmark.
Overall
#33 of 38· High effort· Artificial Analysis
- Vals IndexNot tested
- Arena · Text1425 ±10
#30 of 33· 3,893 votes· Arena
Coding
#31 of 38· High effort· Artificial Analysis
- SciCode44.9%
#32 of 38· High effort· Artificial Analysis
- Arena · WebDev1355 ±16
#33 of 34· 1,659 votes· Arena
- Vibe Code BenchNot tested
- Code MigrationNot tested
Agents & knowledge work
- GDPval-AA14.5
#33 of 39· High effort· Artificial Analysis
- Arena · AgentNot tested
- Finance Agent v2Not tested
- Excel Modeling BenchmarkNot tested
- Legal Research BenchNot tested
- Tax Agent BenchNot tested
Reasoning & science
- Humanity's Last Exam22.0%
#33 of 39· High effort· Artificial Analysis
- CritPt2.6%
#32 of 39· High effort· Artificial Analysis
- ARC-AGI-2saturatingNot tested
- ARC-AGI-3Not tested
- GPQA Diamondsaturating83.5%
#22 of 27· High effort· Artificial Analysis
Knowledge & reliability
#35 of 39· High effort· Artificial Analysis
- AA-LCR83.3%
#7 of 39· High effort· Artificial Analysis
Vision
- MMMU-Pro74.3%
#19 of 27· High effort· Artificial Analysis
- Arena · VisionNot tested
Reported by Meta
Numbers the lab published at launch, run on its own harness and settings. Useful context, but not independently verified and never used in the rankings on this site.
- MCP Atlas (public)75.5%
- OSWorld-Verified65.9%
- SWE-bench Pro51.2%
Specifications
- Lab
- Meta
- Released
- Aug 10, 2026
- Availability
- Generally available
- Context window
- 131,072 tokens
- Max output
- —
- Accepts
- Text, Images
- Knowledge cutoff
- Jan 2026
- Weights
- Open
- Licence
- Apache 2.0
- Parameters
- 29.6B
- Reasoning control
- Effort low to xhigh
Pricing
- Input / 1M tokens
- $0.33
- Cached input / 1M
- $0.04
- Output / 1M tokens
- $1.35
- Blended (3:1)
- $0.58
- Measured cost / task
- $0.06
Typical price at third-party hosts. The lab does not sell paid API access to this model.
Official sources
Closest alternatives
The models nearest to this one on overall score.
23 benchmarks tracked. See methodology and sources.