Model benchmarks
Muse-Glimmer-30B-4bit local LLM performance
As of November 2026, Muse-Glimmer-30B-4bit runs at up to 31.3 tok/s for local inference (best of 6 community benchmark runs across 1 GPU).
Model size
30B
Peak speed
31.3 tok/s
Average speed
30.5 tok/s
Avg PP
632.0 tok/s
Min memory
18.5 GB
Max context
131,072 tokens
Avg output / run
15,199 tokens
Avg runtime / run
8m 31s
Avg quality
82.2
Benchmark runs
6
GPUs tested
1
Quality by task
Average LLM-judged quality (0–100) with the run-to-run spread shown as a P5–P95 band, overall and for each benchmark task, across all 6 runs. The low and high columns show how much the judge’s score varies between runs, and need at least two runs to display.
| Task | P5 (low) | Avg | P95 (high) |
|---|---|---|---|
| Overall | 79.1 | 82.2 | 86.1 |
| Agent Workflow | 74.9 | 86.4 | 91.9 |
| Code Generation | 63.9 | 70.1 | 76.2 |
| Role Play & Narrative | 84.4 | 89.1 | 91.8 |
| Research & Analysis | 78.8 | 83.0 | 86.0 |
Performance by hardware and tool
Every hardware/tool/quantization combination Muse-Glimmer-30B-4bit has been benchmarked on, ranked by peak token generation speed. Last updated November 2026.
| Hardware | Tool | Quant | Best tok/s | Avg tok/s | Memory | Context | Quality | Runs |
|---|---|---|---|---|---|---|---|---|
| Apple M5 Max | oMLX | 4bit | 31.3 tok/s | 31.1 tok/s | 18.5 GB | 131,072 tokens | 82.2 | 3 |
| Apple M5 Max | oMLX | — | 30.7 tok/s | 29.9 tok/s | 18.8 GB | 65,536 tokens | 82.1 | 3 |
Benchmark runs
All 6 Muse-Glimmer-30B-4bit runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is Muse-Glimmer-30B-4bit good for coding?
- In our benchmarks, Muse-Glimmer-30B-4bit scores 70.1/100 for coding. It runs at about 30.5 tok/s, so if you want more speed, Huihui-Qwen3.8-27B-abliterated-Swift-1.5-GSQ-RCO-IQ3_S-mtp is faster (~43.5 tok/s) and still scores well for coding (86.6/100).
- Is Muse-Glimmer-30B-4bit good for agentic (tool-using) tasks?
- In our benchmarks, Muse-Glimmer-30B-4bit scores 86.4/100 for agentic workflows. It runs at about 30.5 tok/s, so if you want more speed, Nex-N2.5-mini-APEX-Mini is faster (~88.2 tok/s) and still scores well for agentic workflows (91.3/100).
- How fast is Muse-Glimmer-30B-4bit for local inference?
- Across 6 community benchmark runs, Muse-Glimmer-30B-4bit reaches up to 31.3 tok/s and averages 30.5 tok/s, with the fastest results on Apple M5 Max.
- How much memory does Muse-Glimmer-30B-4bit need?
- The leanest observed configuration used about 18.5 GB of memory (quantizations tested: 4bit).
- Which tools have been used to run Muse-Glimmer-30B-4bit?
- Benchmarks were submitted using oMLX. Results are community-contributed and updated as new runs arrive.