Model benchmarks

Qwen3.8-27B-oQ2e-mtp local LLM performance

As of August 2026, Qwen3.8-27B-oQ2e-mtp runs at up to 46.3 tok/s for local inference (best of 3 community benchmark runs across 1 GPU).

oMLX
ShareRedditX

Model size

27B

Peak speed

46.3 tok/s

Average speed

44.0 tok/s

Min memory

12.1 GB

Max context

262,144 tokens

Avg output / run

35,136 tokens

Avg runtime / run

23m 20s

Avg quality

10.3

Benchmark runs

3

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 3 runs. The low and high columns show how much the judge’s score varies between runs, and need at least two runs to display.

TaskP5 (low)AvgP95 (high)
Overall4.210.316.2
Agent Workflow10.428.052.5
Code Generation0.00.00.0
Role Play & Narrative3.17.711.8
Research & Analysis2.25.610.8

Performance by hardware and tool

Every hardware/tool/quantization combination Qwen3.8-27B-oQ2e-mtp has been benchmarked on, ranked by peak token generation speed. Last updated August 2026.

HardwareToolQuantBest tok/sAvg tok/sMemoryContextQualityRuns
Apple M5 MaxoMLX46.3 tok/s44.0 tok/s12.1 GB262,144 tokens10.33

Frequently asked questions

Is Qwen3.8-27B-oQ2e-mtp good for coding?
In our benchmarks, Qwen3.8-27B-oQ2e-mtp scores 0.0/100 for coding. It runs at about 44.0 tok/s, so if you want more speed, Gemma4:E2B/QAT-MTP@131K is faster (~303.9 tok/s) and still scores well for coding (66.9/100).
Is Qwen3.8-27B-oQ2e-mtp good for agentic (tool-using) tasks?
In our benchmarks, Qwen3.8-27B-oQ2e-mtp scores 28.0/100 for agentic workflows. It runs at about 44.0 tok/s, so if you want more speed, Gemma4:E2B/QAT-MTP@131K is faster (~303.9 tok/s) and still scores well for agentic workflows (71.7/100).
How fast is Qwen3.8-27B-oQ2e-mtp for local inference?
Across 3 community benchmark runs, Qwen3.8-27B-oQ2e-mtp reaches up to 46.3 tok/s and averages 44.0 tok/s, with the fastest results on Apple M5 Max.
How much memory does Qwen3.8-27B-oQ2e-mtp need?
The leanest observed configuration used about 12.1 GB of memory.
Which tools have been used to run Qwen3.8-27B-oQ2e-mtp?
Benchmarks were submitted using oMLX. Results are community-contributed and updated as new runs arrive.