Model benchmarks

unsloth/Qwen3.8-27B-UD-Q2_K_XL local LLM performance

As of November 2026, unsloth/Qwen3.8-27B-UD-Q2_K_XL runs at up to 50.6 tok/s for local inference (best of 6 community benchmark runs across 1 GPU).

llama.cppQ2_K
ShareRedditX

Model size

27B

Peak speed

50.6 tok/s

Average speed

46.1 tok/s

Avg PP

419.1 tok/s

Min memory

13.9 GB

Max context

65,536 tokens

Avg output / run

71,287 tokens

Avg runtime / run

26m 39s

Avg quality

85.8

Benchmark runs

6

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 6 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)
Overall76.785.890.1
Agent Workflow53.483.195.0
Code Generation80.683.786.5
Role Play & Narrative80.288.994.5
Research & Analysis85.587.589.7

Performance by hardware and tool

Every hardware/tool/quantization combination unsloth/Qwen3.8-27B-UD-Q2_K_XL has been benchmarked on, ranked by peak token generation speed. Last updated November 2026.

HardwareToolQuantBest tok/sAvg tok/sMemoryContextQualityRuns
AMD Radeon RX 7800 XTllama.cppQ2_K50.6 tok/s46.1 tok/s13.9 GB65,536 tokens85.86

Benchmark runs

All 6 unsloth/Qwen3.8-27B-UD-Q2_K_XL runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.

Frequently asked questions

Is unsloth/Qwen3.8-27B-UD-Q2_K_XL good for coding?
In our benchmarks, unsloth/Qwen3.8-27B-UD-Q2_K_XL scores 83.7/100 for coding, among the top 3 for coding on consumer hardware (≤24 GB VRAM). It runs at about 46.1 tok/s, so if you want more speed, ornith-1.5-35b-a3b-apex-mtp-i-quality is faster (~105.2 tok/s) and still scores well for coding (83.6/100).
Is unsloth/Qwen3.8-27B-UD-Q2_K_XL good for agentic (tool-using) tasks?
In our benchmarks, unsloth/Qwen3.8-27B-UD-Q2_K_XL scores 83.1/100 for agentic workflows. It runs at about 46.1 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 unsloth/Qwen3.8-27B-UD-Q2_K_XL for local inference?
Across 6 community benchmark runs, unsloth/Qwen3.8-27B-UD-Q2_K_XL reaches up to 50.6 tok/s and averages 46.1 tok/s, with the fastest results on AMD Radeon RX 7800 XT.
How much memory does unsloth/Qwen3.8-27B-UD-Q2_K_XL need?
The leanest observed configuration used about 13.9 GB of memory (quantizations tested: Q2_K).
Which tools have been used to run unsloth/Qwen3.8-27B-UD-Q2_K_XL?
Benchmarks were submitted using llama.cpp. Results are community-contributed and updated as new runs arrive.