Model benchmarks
huihui-qwen3.8-27b-abliterated-oq5e-mtp local LLM performance
As of September 2026, huihui-qwen3.8-27b-abliterated-oq5e-mtp runs at up to 32.2 tok/s for local inference (best of 2 community benchmark runs across 1 GPU).
Model size
27B
Peak speed
32.2 tok/s
Average speed
30.9 tok/s
Avg prefill
604.6 tok/s
Min memory
18.9 GB
Max context
131,072 tokens
Avg output / run
260 tokens
Avg runtime / run
19s
Avg quality
1.6
Benchmark runs
2
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 2 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 | 0.4 | 1.6 | 2.8 |
| Agent Workflow | 0.5 | 4.9 | 9.3 |
| Code Generation | 0.0 | 0.0 | 0.0 |
| Role Play & Narrative | 0.9 | 1.4 | 1.8 |
| Research & Analysis | 0.0 | 0.0 | 0.0 |
Performance by hardware and tool
Every hardware/tool/quantization combination huihui-qwen3.8-27b-abliterated-oq5e-mtp has been benchmarked on, ranked by peak token generation speed. Last updated September 2026.
| Hardware | Tool | Quant | Best tok/s | Avg tok/s | Memory | Context | Quality | Runs |
|---|---|---|---|---|---|---|---|---|
| Apple M5 Max | LM Studio | — | 32.2 tok/s | 30.9 tok/s | 18.9 GB | 131,072 tokens | 1.6 | 2 |
Frequently asked questions
- Is huihui-qwen3.8-27b-abliterated-oq5e-mtp good for coding?
- In our benchmarks, huihui-qwen3.8-27b-abliterated-oq5e-mtp scores 0.0/100 for coding. It runs at about 30.9 tok/s, so if you want more speed, K2-Horizon-MoVA-36B-A4B-MLX-4bit is faster (~48.8 tok/s) and still scores well for coding (85.3/100).
- Is huihui-qwen3.8-27b-abliterated-oq5e-mtp good for agentic (tool-using) tasks?
- In our benchmarks, huihui-qwen3.8-27b-abliterated-oq5e-mtp scores 4.9/100 for agentic workflows. It runs at about 30.9 tok/s, so if you want more speed, granite-4.2-8b-Q4_K_L is faster (~71.6 tok/s) and still scores well for agentic workflows (92.9/100).
- How fast is huihui-qwen3.8-27b-abliterated-oq5e-mtp for local inference?
- Across 2 community benchmark runs, huihui-qwen3.8-27b-abliterated-oq5e-mtp reaches up to 32.2 tok/s and averages 30.9 tok/s, with the fastest results on Apple M5 Max.
- How much memory does huihui-qwen3.8-27b-abliterated-oq5e-mtp need?
- The leanest observed configuration used about 18.9 GB of memory.
- Which tools have been used to run huihui-qwen3.8-27b-abliterated-oq5e-mtp?
- Benchmarks were submitted using LM Studio. Results are community-contributed and updated as new runs arrive.