Is Qwen3.8-27B-UD-IQ4_XS good for coding?

In our benchmarks, Qwen3.8-27B-UD-IQ4_XS scores 74.1/100 for coding. It runs at about 49.2 tok/s, so if you want more speed, local-qwen is faster (~60.2 tok/s) and still scores well for coding (84.8/100).

Is Qwen3.8-27B-UD-IQ4_XS good for agentic (tool-using) tasks?

In our benchmarks, Qwen3.8-27B-UD-IQ4_XS scores 85.4/100 for agentic workflows. It runs at about 49.2 tok/s, so if you want more speed, granite-4.2-8b-Q4_K_L is faster (~71.6 tok/s) and still scores well for agentic workflows (92.9/100).

How fast is Qwen3.8-27B-UD-IQ4_XS for local inference?

Across 9 community benchmark runs, Qwen3.8-27B-UD-IQ4_XS reaches up to 75.0 tok/s and averages 49.2 tok/s, with the fastest results on NVIDIA GeForce RTX 2060.

Which tools have been used to run Qwen3.8-27B-UD-IQ4_XS?

Benchmarks were submitted using llama.cpp, vLLM. Results are community-contributed and updated as new runs arrive.

Model benchmarks

Qwen3.8-27B-UD-IQ4_XS local LLM performance

As of October 2026, Qwen3.8-27B-UD-IQ4_XS runs at up to 75.0 tok/s for local inference (best of 9 community benchmark runs across 5 GPUs).

llama.cppvLLM

ShareReddit X

Model size

27B

Peak speed

75.0 tok/s

Average speed

49.2 tok/s

Avg PP

680.8 tok/s

Min memory

n/a

Max context

262,144 tokens

Avg output / run

28,412 tokens

Avg runtime / run

11m 10s

Avg quality

83.4

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 9 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	79.0	83.4	86.8
Agent Workflow	80.1	85.4	90.8
Code Generation	55.6	74.1	83.1
Role Play & Narrative	80.1	87.6	91.5
Research & Analysis	82.4	86.4	89.5

Performance by hardware and tool

Every hardware/tool/quantization combination Qwen3.8-27B-UD-IQ4_XS 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 2060	llama.cpp	—	75.0 tok/s	70.6 tok/s	n/a	196,608 tokens	86.6	2
NVIDIA GeForce RTX 3090	llama.cpp	—	74.0 tok/s	65.0 tok/s	n/a	262,144 tokens	83.0	2
NVIDIA GeForce RTX 5070 Ti	llama.cpp	—	46.8 tok/s	46.8 tok/s	n/a	32,768 tokens	78.9	1
NVIDIA GeForce RTX 5060	llama.cpp	—	38.1 tok/s	31.6 tok/s	n/a	65,536 tokens	81.9	2
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

Benchmark runs

All 9 Qwen3.8-27B-UD-IQ4_XS 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-UD-IQ4_XS good for coding?: In our benchmarks, Qwen3.8-27B-UD-IQ4_XS scores 74.1/100 for coding. It runs at about 49.2 tok/s, so if you want more speed, local-qwen is faster (~60.2 tok/s) and still scores well for coding (84.8/100).
Is Qwen3.8-27B-UD-IQ4_XS good for agentic (tool-using) tasks?: In our benchmarks, Qwen3.8-27B-UD-IQ4_XS scores 85.4/100 for agentic workflows. It runs at about 49.2 tok/s, so if you want more speed, granite-4.2-8b-Q4_K_L is faster (~71.6 tok/s) and still scores well for agentic workflows (92.9/100).
How fast is Qwen3.8-27B-UD-IQ4_XS for local inference?: Across 9 community benchmark runs, Qwen3.8-27B-UD-IQ4_XS reaches up to 75.0 tok/s and averages 49.2 tok/s, with the fastest results on NVIDIA GeForce RTX 2060.
Which tools have been used to run Qwen3.8-27B-UD-IQ4_XS?: Benchmarks were submitted using llama.cpp, vLLM. Results are community-contributed and updated as new runs arrive.