Is mtplx-qwen38-27b-optimized-quality good for coding?

In our benchmarks, mtplx-qwen38-27b-optimized-quality scores 77.5/100 for coding. It runs at about 33.2 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 mtplx-qwen38-27b-optimized-quality good for agentic (tool-using) tasks?

In our benchmarks, mtplx-qwen38-27b-optimized-quality scores 85.9/100 for agentic workflows. It runs at about 33.2 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 mtplx-qwen38-27b-optimized-quality for local inference?

Across 7 community benchmark runs, mtplx-qwen38-27b-optimized-quality reaches up to 37.8 tok/s and averages 33.2 tok/s, with the fastest results on Apple M5 Max.

Which tools have been used to run mtplx-qwen38-27b-optimized-quality?

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

Model benchmarks

mtplx-qwen38-27b-optimized-quality local LLM performance

As of September 2026, mtplx-qwen38-27b-optimized-quality runs at up to 37.8 tok/s for local inference (best of 7 community benchmark runs across 3 GPUs).

vLLM

ShareReddit X

Model size

27B

Peak speed

37.8 tok/s

Average speed

33.2 tok/s

Min memory

n/a

Max context

128,000 tokens

Avg output / run

54,736 tokens

Avg runtime / run

24m 15s

Avg quality

85.6

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 7 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	81.5	85.6	88.7
Agent Workflow	76.2	85.9	91.2
Code Generation	70.4	77.5	84.3
Role Play & Narrative	87.9	92.5	96.1
Research & Analysis	83.6	86.5	88.7

Performance by hardware and tool

Every hardware/tool/quantization combination mtplx-qwen38-27b-optimized-quality 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
Apple M5 Max	vLLM	—	37.8 tok/s	35.0 tok/s	n/a	128,000 tokens	87.2	3
Apple M4 Max	vLLM	—	36.2 tok/s	35.5 tok/s	n/a	128,000 tokens	85.7	3
Apple M5 Pro	vLLM	—	21.0 tok/s	21.0 tok/s	n/a	8,192 tokens	80.5	1

Benchmark runs

All 7 mtplx-qwen38-27b-optimized-quality runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.

Open run page →

Frequently asked questions

Is mtplx-qwen38-27b-optimized-quality good for coding?: In our benchmarks, mtplx-qwen38-27b-optimized-quality scores 77.5/100 for coding. It runs at about 33.2 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 mtplx-qwen38-27b-optimized-quality good for agentic (tool-using) tasks?: In our benchmarks, mtplx-qwen38-27b-optimized-quality scores 85.9/100 for agentic workflows. It runs at about 33.2 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 mtplx-qwen38-27b-optimized-quality for local inference?: Across 7 community benchmark runs, mtplx-qwen38-27b-optimized-quality reaches up to 37.8 tok/s and averages 33.2 tok/s, with the fastest results on Apple M5 Max.
Which tools have been used to run mtplx-qwen38-27b-optimized-quality?: Benchmarks were submitted using vLLM. Results are community-contributed and updated as new runs arrive.