Model benchmarks
cyankiwi/Qwen3.8-27B-AWQ-INT4 local LLM performance
As of October 2026, cyankiwi/Qwen3.8-27B-AWQ-INT4 runs at up to 185.0 tok/s for local inference (best of 2 community benchmark runs across 1 GPU).
Model size
27B
Peak speed
185.0 tok/s
Average speed
111.1 tok/s
Avg PP
538.5 tok/s
Min memory
31.7 GB
Max context
262,144 tokens
Avg output / run
71,515 tokens
Avg runtime / run
35m 14s
Avg quality
88.1
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 | 86.7 | 88.1 | 89.4 |
| Agent Workflow | 89.2 | 90.3 | 91.4 |
| Code Generation | 77.3 | 83.2 | 89.0 |
| Role Play & Narrative | 91.7 | 93.0 | 94.3 |
| Research & Analysis | 82.9 | 85.7 | 88.5 |
Performance by hardware and tool
Every hardware/tool/quantization combination cyankiwi/Qwen3.8-27B-AWQ-INT4 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 |
|---|---|---|---|---|---|---|---|---|
| CPU only | vLLM | INT4 | 185.0 tok/s | 185.0 tok/s | n/a | 65,536 tokens | 86.5 | 1 |
| NVIDIA GeForce RTX 3080 Ti | sglang | INT4 | 37.2 tok/s | 37.2 tok/s | 31.7 GB | 262,144 tokens | 89.6 | 1 |
Benchmark runs
All 2 cyankiwi/Qwen3.8-27B-AWQ-INT4 runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is ulkaa/Qwen3.8-27B-AWQ-INT4 good for coding?
- In our benchmarks, ulkaa/Qwen3.8-27B-AWQ-INT4 scores 83.2/100 for coding. It runs at about 111.1 tok/s, so if you want more speed, Qwen3.6-35B-A3B-Q4_K_XL is faster (~156.9 tok/s) and still scores well for coding (82.2/100).
- Is ulkaa/Qwen3.8-27B-AWQ-INT4 good for agentic (tool-using) tasks?
- In our benchmarks, ulkaa/Qwen3.8-27B-AWQ-INT4 scores 90.3/100 for agentic workflows. It runs at about 111.1 tok/s, so if you want more speed, Tiel-Coder-35B-A3B-Q4_K_S is faster (~162.4 tok/s) and still scores well for agentic workflows (87.8/100).
- How fast is cyankiwi/Qwen3.8-27B-AWQ-INT4 for local inference?
- Across 2 community benchmark runs, cyankiwi/Qwen3.8-27B-AWQ-INT4 reaches up to 185.0 tok/s and averages 111.1 tok/s.
- How much memory does cyankiwi/Qwen3.8-27B-AWQ-INT4 need?
- The leanest observed configuration used about 31.7 GB of memory (quantizations tested: INT4).
- Which tools have been used to run cyankiwi/Qwen3.8-27B-AWQ-INT4?
- Benchmarks were submitted using sglang, vLLM. Results are community-contributed and updated as new runs arrive.