Model benchmarks
Dirk-Qwen3.8-27B-GSQ-RCO-IQ3_S local LLM performance
As of September 2026, Dirk-Qwen3.8-27B-GSQ-RCO-IQ3_S runs at up to 21.4 tok/s for local inference (best of 2 community benchmark runs across 1 GPU).
Model size
27B
Peak speed
21.4 tok/s
Average speed
21.4 tok/s
Avg prefill
209.9 tok/s
Min memory
13.2 GB
Max context
145,152 tokens
Avg output / run
19,506 tokens
Avg runtime / run
14m 51s
Avg quality
84.5
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 | 83.3 | 84.5 | 85.7 |
| Agent Workflow | 81.3 | 83.6 | 85.8 |
| Code Generation | 75.2 | 76.3 | 77.3 |
| Role Play & Narrative | 90.2 | 90.5 | 90.8 |
| Research & Analysis | 86.4 | 87.7 | 88.9 |
Performance by hardware and tool
Every hardware/tool/quantization combination Dirk-Qwen3.8-27B-GSQ-RCO-IQ3_S 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 |
|---|---|---|---|---|---|---|---|---|
| AMD Radeon RX 5600 OEM/5600 XT / 5700/5700 XT | llama.cpp | — | 21.4 tok/s | 21.4 tok/s | 13.2 GB | 145,152 tokens | 84.5 | 2 |
Frequently asked questions
- Is Dirk-Qwen3.8-27B-GSQ-RCO-IQ3_S good for coding?
- In our benchmarks, Dirk-Qwen3.8-27B-GSQ-RCO-IQ3_S scores 76.3/100 for coding. It runs at about 21.4 tok/s, so if you want more speed, Qwen3.6-35B-A3B-IQ4_NL is faster (~165.9 tok/s) and still scores well for coding (68.7/100).
- Is Dirk-Qwen3.8-27B-GSQ-RCO-IQ3_S good for agentic (tool-using) tasks?
- In our benchmarks, Dirk-Qwen3.8-27B-GSQ-RCO-IQ3_S scores 83.6/100 for agentic workflows. It runs at about 21.4 tok/s, so if you want more speed, unsloth/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF:UD-Q4_K_XL is faster (~169.2 tok/s) and still scores well for agentic workflows (83.0/100).
- How fast is Dirk-Qwen3.8-27B-GSQ-RCO-IQ3_S for local inference?
- Across 2 community benchmark runs, Dirk-Qwen3.8-27B-GSQ-RCO-IQ3_S reaches up to 21.4 tok/s and averages 21.4 tok/s, with the fastest results on AMD Radeon RX 5600 OEM/5600 XT / 5700/5700 XT.
- How much memory does Dirk-Qwen3.8-27B-GSQ-RCO-IQ3_S need?
- The leanest observed configuration used about 13.2 GB of memory.
- Which tools have been used to run Dirk-Qwen3.8-27B-GSQ-RCO-IQ3_S?
- Benchmarks were submitted using llama.cpp. Results are community-contributed and updated as new runs arrive.