Is Qwen3.8-27B-4bit good for coding?

In our benchmarks, Qwen3.8-27B-4bit scores 73.5/100 for coding. It runs at about 60.0 tok/s, so if you want more speed, ornith-1.5-35b-a3b-apex-i-quality is faster (~75.9 tok/s) and still scores well for coding (82.9/100).

Is Qwen3.8-27B-4bit good for agentic (tool-using) tasks?

In our benchmarks, Qwen3.8-27B-4bit scores 78.0/100 for agentic workflows. It runs at about 60.0 tok/s, so if you want more speed, Nex-N2.5-mini-APEX-Mini is faster (~88.2 tok/s) and still scores well for agentic workflows (91.3/100).

How fast is Qwen3.8-27B-4bit for local inference?

Across 22 community benchmark runs, Qwen3.8-27B-4bit reaches up to 134.8 tok/s and averages 60.0 tok/s, with the fastest results on Apple M5 Ultra.

How much memory does Qwen3.8-27B-4bit need?

The leanest observed configuration used about 17.1 GB of memory (quantizations tested: 4bit).

Which tools have been used to run Qwen3.8-27B-4bit?

Benchmarks were submitted using OpenAI-compatible, oMLX, splash, vLLM. Results are community-contributed and updated as new runs arrive.

Model benchmarks

Qwen3.8-27B-4bit local LLM performance

As of November 2026, Qwen3.8-27B-4bit runs at up to 134.8 tok/s for local inference (best of 22 community benchmark runs across 3 GPUs).

OpenAI-compatibleoMLXsplashvLLM4bit

ShareReddit X

Model size

27B

Peak speed

134.8 tok/s

Average speed

60.0 tok/s

Avg PP

796.4 tok/s

Min memory

17.1 GB

Max context

262,144 tokens

Avg output / run

50,625 tokens

Avg runtime / run

12m 51s

Avg quality

81.1

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 22 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	67.6	81.1	88.7
Agent Workflow	58.8	78.0	91.6
Code Generation	65.1	73.5	81.8
Role Play & Narrative	78.4	89.2	95.5
Research & Analysis	74.4	83.9	90.3

Performance by hardware and tool

Every hardware/tool/quantization combination Qwen3.8-27B-4bit has been benchmarked on, ranked by peak token generation speed. Last updated November 2026.

Hardware	Tool	Quant	Best tok/s	Avg tok/s	Memory	Context	Quality	Runs
Apple M5 Ultra	splash	—	134.8 tok/s	128.0 tok/s	20.2 GB	65,536 tokens	81.6	2
Apple M5 Max	splash	—	99.4 tok/s	95.6 tok/s	20.0 GB	65,536 tokens	82.1	3
Apple M5 Max	vLLM	—	90.5 tok/s	82.9 tok/s	n/a	131,072 tokens	84.6	4
Apple M5 Max	OpenAI-compatible	—	84.4 tok/s	84.4 tok/s	21.4 GB	65,536 tokens	86.9	1
Apple M4 Pro	oMLX	4bit	40.0 tok/s	28.9 tok/s	17.1 GB	131,072 tokens	71.1	3
Apple M5 Max	oMLX	—	31.9 tok/s	30.6 tok/s	17.8 GB	262,144 tokens	81.9	9

Benchmark runs

All 22 Qwen3.8-27B-4bit 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-4bit good for coding?: In our benchmarks, Qwen3.8-27B-4bit scores 73.5/100 for coding. It runs at about 60.0 tok/s, so if you want more speed, ornith-1.5-35b-a3b-apex-i-quality is faster (~75.9 tok/s) and still scores well for coding (82.9/100).
Is Qwen3.8-27B-4bit good for agentic (tool-using) tasks?: In our benchmarks, Qwen3.8-27B-4bit scores 78.0/100 for agentic workflows. It runs at about 60.0 tok/s, so if you want more speed, Nex-N2.5-mini-APEX-Mini is faster (~88.2 tok/s) and still scores well for agentic workflows (91.3/100).
How fast is Qwen3.8-27B-4bit for local inference?: Across 22 community benchmark runs, Qwen3.8-27B-4bit reaches up to 134.8 tok/s and averages 60.0 tok/s, with the fastest results on Apple M5 Ultra.
How much memory does Qwen3.8-27B-4bit need?: The leanest observed configuration used about 17.1 GB of memory (quantizations tested: 4bit).
Which tools have been used to run Qwen3.8-27B-4bit?: Benchmarks were submitted using OpenAI-compatible, oMLX, splash, vLLM. Results are community-contributed and updated as new runs arrive.