Model benchmarks
tensorfold-qwen38-27b local LLM performance
As of October 2026, tensorfold-qwen38-27b runs at up to 25.1 tok/s for local inference (best of 2 community benchmark runs across 1 GPU).
Model size
27B
Peak speed
25.1 tok/s
Average speed
25.0 tok/s
Avg PP
76.3 tok/s
Min memory
22.4 GB
Max context
16,384 tokens
Avg output / run
18,652 tokens
Avg runtime / run
12m 38s
Avg quality
84.8
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 | 84.7 | 84.8 | 84.9 |
| Agent Workflow | 79.8 | 79.9 | 80.1 |
| Code Generation | 84.2 | 84.3 | 84.4 |
| Role Play & Narrative | 90.5 | 91.8 | 93.2 |
| Research & Analysis | 81.9 | 83.1 | 84.3 |
Performance by hardware and tool
Every hardware/tool/quantization combination tensorfold-qwen38-27b 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 |
|---|---|---|---|---|---|---|---|---|
| Apple M4 Pro | TensorFold | — | 25.1 tok/s | 25.0 tok/s | 22.4 GB | 16,384 tokens | 84.8 | 2 |
Benchmark runs
All 2 tensorfold-qwen38-27b runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is tensorfold-qwen38-27b good for coding?
- In our benchmarks, tensorfold-qwen38-27b scores 84.3/100 for coding, currently the best-scoring model for coding that runs on consumer hardware (≤24 GB VRAM). It runs at about 25.0 tok/s, so if you want more speed, Qwen3.8-27B-Turbo-Brilliance-50X-RI-GSQ-RCO-IQ3_XXS-mtp is faster (~37.5 tok/s) and still scores well for coding (83.8/100).
- Is tensorfold-qwen38-27b good for agentic (tool-using) tasks?
- In our benchmarks, tensorfold-qwen38-27b scores 79.9/100 for agentic workflows. It runs at about 25.0 tok/s, so if you want more speed, muse-glimmer:latest is faster (~33.3 tok/s) and still scores well for agentic workflows (91.6/100).
- How fast is tensorfold-qwen38-27b for local inference?
- Across 2 community benchmark runs, tensorfold-qwen38-27b reaches up to 25.1 tok/s and averages 25.0 tok/s, with the fastest results on Apple M4 Pro.
- How much memory does tensorfold-qwen38-27b need?
- The leanest observed configuration used about 22.4 GB of memory.
- Which tools have been used to run tensorfold-qwen38-27b?
- Benchmarks were submitted using TensorFold. Results are community-contributed and updated as new runs arrive.