Is qwen3.8:27b-q4_K_M good for coding?

In our benchmarks, qwen3.8:27b-q4_K_M scores 79.2/100 for coding. It runs at about 28.5 tok/s, so if you want more speed, Huihui-Qwen3.8-27B-abliterated-Swift-1.5-GSQ-RCO-IQ3_S-mtp is faster (~43.5 tok/s) and still scores well for coding (86.6/100).

Is qwen3.8:27b-q4_K_M good for agentic (tool-using) tasks?

In our benchmarks, qwen3.8:27b-q4_K_M scores 81.9/100 for agentic workflows. It runs at about 28.5 tok/s, so if you want more speed, muse-glimmer:latest is faster (~33.3 tok/s) and still scores well for agentic workflows (91.6/100).

How fast is qwen3.8:27b-q4_K_M for local inference?

Across 8 community benchmark runs, qwen3.8:27b-q4_K_M reaches up to 35.8 tok/s and averages 28.5 tok/s, with the fastest results on AMD Radeon RX 7900 XTX.

How much memory does qwen3.8:27b-q4_K_M need?

The leanest observed configuration used about 19.3 GB of memory (quantizations tested: Q4_K_M).

Which tools have been used to run qwen3.8:27b-q4_K_M?

Benchmarks were submitted using Ollama. Results are community-contributed and updated as new runs arrive.

Model benchmarks

qwen3.8:27b-q4_K_M local LLM performance

As of October 2026, qwen3.8:27b-q4_K_M runs at up to 35.8 tok/s for local inference (best of 8 community benchmark runs across 2 GPUs).

OllamaQ4_K_M

ShareReddit X

Model size

27.3B

Peak speed

35.8 tok/s

Average speed

28.5 tok/s

Avg PP

465.3 tok/s

Min memory

19.3 GB

Max context

128,000 tokens

Avg output / run

14,987 tokens

Avg runtime / run

9m 28s

Avg quality

84.4

Benchmark runs

GPUs tested

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 8 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	80.4	84.4	88.3
Agent Workflow	73.2	81.9	90.0
Code Generation	73.3	79.2	83.5
Role Play & Narrative	86.0	91.3	94.2
Research & Analysis	81.4	85.3	88.4

Performance by hardware and tool

Every hardware/tool/quantization combination qwen3.8:27b-q4_K_M 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
AMD Radeon RX 7900 XTX	Ollama	Q4_K_M	35.8 tok/s	34.5 tok/s	19.3 GB	65,536 tokens	85.8	2
AMD Radeon RX 7800 XT	Ollama	Q4_K_M	27.1 tok/s	26.4 tok/s	19.5 GB	128,000 tokens	83.9	6

Benchmark runs

All 8 qwen3.8:27b-q4_K_M runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.

Open run page →

Frequently asked questions

Is qwen3.8:27b-q4_K_M good for coding?: In our benchmarks, qwen3.8:27b-q4_K_M scores 79.2/100 for coding. It runs at about 28.5 tok/s, so if you want more speed, Huihui-Qwen3.8-27B-abliterated-Swift-1.5-GSQ-RCO-IQ3_S-mtp is faster (~43.5 tok/s) and still scores well for coding (86.6/100).
Is qwen3.8:27b-q4_K_M good for agentic (tool-using) tasks?: In our benchmarks, qwen3.8:27b-q4_K_M scores 81.9/100 for agentic workflows. It runs at about 28.5 tok/s, so if you want more speed, muse-glimmer:latest is faster (~33.3 tok/s) and still scores well for agentic workflows (91.6/100).
How fast is qwen3.8:27b-q4_K_M for local inference?: Across 8 community benchmark runs, qwen3.8:27b-q4_K_M reaches up to 35.8 tok/s and averages 28.5 tok/s, with the fastest results on AMD Radeon RX 7900 XTX.
How much memory does qwen3.8:27b-q4_K_M need?: The leanest observed configuration used about 19.3 GB of memory (quantizations tested: Q4_K_M).
Which tools have been used to run qwen3.8:27b-q4_K_M?: Benchmarks were submitted using Ollama. Results are community-contributed and updated as new runs arrive.