Model benchmarks
Arete-27B_t1p95k20m0 local LLM performance
As of November 2026, Arete-27B_t1p95k20m0 runs at up to 37.2 tok/s for local inference (best of 2 community benchmark runs across 1 GPU).
Model size
27B
Peak speed
37.2 tok/s
Average speed
37.2 tok/s
Avg PP
694.1 tok/s
Min memory
52.2 GB
Max context
32,768 tokens
Avg output / run
40,330 tokens
Avg runtime / run
18m 13s
Avg quality
86.5
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 | 86.4 | 86.5 | 86.6 |
| Agent Workflow | 86.4 | 87.7 | 89.1 |
| Code Generation | 75.0 | 78.8 | 82.6 |
| Role Play & Narrative | 90.4 | 91.9 | 93.4 |
| Research & Analysis | 87.2 | 87.8 | 88.3 |
Performance by hardware and tool
Every hardware/tool/quantization combination Arete-27B_t1p95k20m0 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 |
|---|---|---|---|---|---|---|---|---|
| Tesla V100-PCIE-32GB | llama.cpp | — | 37.2 tok/s | 37.2 tok/s | 52.2 GB | 32,768 tokens | 86.5 | 2 |
Benchmark runs
All 2 Arete-27B_t1p95k20m0 runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is Arete-27B_t1p95k20m0 good for coding?
- In our benchmarks, Arete-27B_t1p95k20m0 scores 78.8/100 for coding. It runs at about 37.2 tok/s, so if you want more speed, local-qwen is faster (~60.2 tok/s) and still scores well for coding (84.8/100).
- Is Arete-27B_t1p95k20m0 good for agentic (tool-using) tasks?
- In our benchmarks, Arete-27B_t1p95k20m0 scores 87.7/100 for agentic workflows. It runs at about 37.2 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 Arete-27B_t1p95k20m0 for local inference?
- Across 2 community benchmark runs, Arete-27B_t1p95k20m0 reaches up to 37.2 tok/s and averages 37.2 tok/s, with the fastest results on Tesla V100-PCIE-32GB.
- How much memory does Arete-27B_t1p95k20m0 need?
- The leanest observed configuration used about 52.2 GB of memory.
- Which tools have been used to run Arete-27B_t1p95k20m0?
- Benchmarks were submitted using llama.cpp. Results are community-contributed and updated as new runs arrive.