Model benchmarks
unsloth/Muse-Glimmer-30B-GGUF:UD-Q4_K_XL local LLM performance
As of October 2026, unsloth/Muse-Glimmer-30B-GGUF:UD-Q4_K_XL runs at up to 68.1 tok/s for local inference (best of 8 community benchmark runs across 4 GPUs).
Model size
30B
Peak speed
68.1 tok/s
Average speed
38.1 tok/s
Avg PP
558.6 tok/s
Min memory
17.5 GB
Max context
131,072 tokens
Avg output / run
16,473 tokens
Avg runtime / run
7m 55s
Avg quality
82.8
Benchmark runs
8
GPUs tested
4
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 8 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 | 80.4 | 82.8 | 85.4 |
| Agent Workflow | 83.5 | 87.0 | 90.1 |
| Code Generation | 67.9 | 74.6 | 79.4 |
| Role Play & Narrative | 81.5 | 85.6 | 90.0 |
| Research & Analysis | 81.8 | 84.0 | 87.6 |
Performance by hardware and tool
Every hardware/tool/quantization combination unsloth/Muse-Glimmer-30B-GGUF:UD-Q4_K_XL has been benchmarked on, ranked by peak token generation speed. Last updated October 2026.
| Hardware | Tool | Quant | Best tok/s | Avg tok/s | Memory | Context | Quality | Runs |
|---|---|---|---|---|---|---|---|---|
| NVIDIA GeForce RTX 3080 Ti | llama.cpp | Q4_K | 68.1 tok/s | 68.1 tok/s | n/a | 65,536 tokens | 81.9 | 1 |
| Advanced Micro Devices, Inc. [AMD/ATI] Device 7551 | llama.cpp | Q4_K | 34.7 tok/s | 34.3 tok/s | n/a | 131,072 tokens | 81.7 | 3 |
| AMD Radeon AI PRO R9700 | llama.cpp | Q4_K | 34.6 tok/s | 34.5 tok/s | 17.5 GB | 131,072 tokens | 84.5 | 3 |
| Apple M5 Max | llama.cpp | Q4_K | 30.8 tok/s | 30.8 tok/s | n/a | 65,536 tokens | 82.0 | 1 |
Benchmark runs
All 8 unsloth/Muse-Glimmer-30B-GGUF:UD-Q4_K_XL runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is unsloth/Muse-Glimmer-30B-GGUF:UD-Q4_K_XL good for coding?
- In our benchmarks, unsloth/Muse-Glimmer-30B-GGUF:UD-Q4_K_XL scores 74.6/100 for coding. It runs at about 38.1 tok/s, so if you want more speed, ornith-1.5-35b-a3b-apex-mtp-i-quality is faster (~105.2 tok/s) and still scores well for coding (83.6/100).
- Is unsloth/Muse-Glimmer-30B-GGUF:UD-Q4_K_XL good for agentic (tool-using) tasks?
- In our benchmarks, unsloth/Muse-Glimmer-30B-GGUF:UD-Q4_K_XL scores 87.0/100 for agentic workflows. It runs at about 38.1 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 unsloth/Muse-Glimmer-30B-GGUF:UD-Q4_K_XL for local inference?
- Across 8 community benchmark runs, unsloth/Muse-Glimmer-30B-GGUF:UD-Q4_K_XL reaches up to 68.1 tok/s and averages 38.1 tok/s, with the fastest results on NVIDIA GeForce RTX 3080 Ti.
- How much memory does unsloth/Muse-Glimmer-30B-GGUF:UD-Q4_K_XL need?
- The leanest observed configuration used about 17.5 GB of memory (quantizations tested: Q4_K).
- Which tools have been used to run unsloth/Muse-Glimmer-30B-GGUF:UD-Q4_K_XL?
- Benchmarks were submitted using llama.cpp. Results are community-contributed and updated as new runs arrive.