Model benchmarks
/models/Qwen3.8-27B-UD-IQ4_XS.gguf local LLM performance
As of August 2026, /models/Qwen3.8-27B-UD-IQ4_XS.gguf runs at up to 30.9 tok/s for local inference (best of 2 community benchmark runs across 1 GPU).
Model size
27B
Peak speed
30.9 tok/s
Average speed
30.8 tok/s
Min memory
n/a
Max context
49,152 tokens
Avg output / run
15,424 tokens
Avg runtime / run
9m 25s
Avg quality
84.1
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 | 83.9 | 84.1 | 84.2 |
| Agent Workflow | 82.8 | 82.9 | 82.9 |
| Code Generation | 77.3 | 77.8 | 78.2 |
| Role Play & Narrative | 88.2 | 89.5 | 90.8 |
| Research & Analysis | 84.6 | 86.2 | 87.8 |
Performance by hardware and tool
Every hardware/tool/quantization combination /models/Qwen3.8-27B-UD-IQ4_XS.gguf has been benchmarked on, ranked by peak token generation speed. Last updated August 2026.
| Hardware | Tool | Quant | Best tok/s | Avg tok/s | Memory | Context | Quality | Runs |
|---|---|---|---|---|---|---|---|---|
| AMD Radeon RX 9070/9070 XT/9070 GRE | vLLM | — | 30.9 tok/s | 30.8 tok/s | n/a | 49,152 tokens | 84.1 | 2 |
Frequently asked questions
- Is /models/Qwen3.8-27B-UD-IQ4_XS.gguf good for coding?
- In our benchmarks, /models/Qwen3.8-27B-UD-IQ4_XS.gguf scores 77.8/100 for coding. It runs at about 30.8 tok/s, so if you want more speed, mudler/Ornith-1.5-35B-A3B-APEX-MTP-GGUF is faster (~190.9 tok/s) and still scores well for coding (73.4/100).
- Is /models/Qwen3.8-27B-UD-IQ4_XS.gguf good for agentic (tool-using) tasks?
- In our benchmarks, /models/Qwen3.8-27B-UD-IQ4_XS.gguf scores 82.9/100 for agentic workflows. It runs at about 30.8 tok/s, so if you want more speed, mudler/Ornith-1.5-35B-A3B-APEX-MTP-GGUF is faster (~190.9 tok/s) and still scores well for agentic workflows (85.7/100).
- How fast is /models/Qwen3.8-27B-UD-IQ4_XS.gguf for local inference?
- Across 2 community benchmark runs, /models/Qwen3.8-27B-UD-IQ4_XS.gguf reaches up to 30.9 tok/s and averages 30.8 tok/s, with the fastest results on AMD Radeon RX 9070/9070 XT/9070 GRE.
- How much memory does /models/Qwen3.8-27B-UD-IQ4_XS.gguf need?
- The leanest observed configuration used about n/a of memory.
- Which tools have been used to run /models/Qwen3.8-27B-UD-IQ4_XS.gguf?
- Benchmarks were submitted using vLLM. Results are community-contributed and updated as new runs arrive.