Model benchmarks
swift-1.5-qwen3.8-27b-uncensored local LLM performance
As of October 2026, swift-1.5-qwen3.8-27b-uncensored runs at up to 302.2 tok/s for local inference (best of 19 community benchmark runs across 1 GPU).
Model size
27B
Peak speed
302.2 tok/s
Average speed
284.1 tok/s
Avg PP
13643.1 tok/s
Min memory
28.8 GB
Max context
120,000 tokens
Avg output / run
20,574 tokens
Avg runtime / run
1m 10s
Avg quality
83.6
Benchmark runs
19
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 19 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 | 83.6 | 86.9 |
| Agent Workflow | 70.7 | 82.3 | 90.8 |
| Code Generation | 69.3 | 76.7 | 85.0 |
| Role Play & Narrative | 85.8 | 91.0 | 95.3 |
| Research & Analysis | 78.6 | 84.3 | 87.8 |
Performance by hardware and tool
Every hardware/tool/quantization combination swift-1.5-qwen3.8-27b-uncensored 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 5060 Ti | ninfer | — | 302.2 tok/s | 284.1 tok/s | 28.8 GB | 120,000 tokens | 83.6 | 19 |
Benchmark runs
All 19 swift-1.5-qwen3.8-27b-uncensored runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- How fast is swift-1.5-qwen3.8-27b-uncensored for local inference?
- Across 19 community benchmark runs, swift-1.5-qwen3.8-27b-uncensored reaches up to 302.2 tok/s and averages 284.1 tok/s, with the fastest results on NVIDIA GeForce RTX 5060 Ti.
- How much memory does swift-1.5-qwen3.8-27b-uncensored need?
- The leanest observed configuration used about 28.8 GB of memory.
- Which tools have been used to run swift-1.5-qwen3.8-27b-uncensored?
- Benchmarks were submitted using ninfer. Results are community-contributed and updated as new runs arrive.