Model benchmarks

unsloth/Qwen3.8-27B-GGUF:UD-IQ3_S local LLM performance

As of September 2026, unsloth/Qwen3.8-27B-GGUF:UD-IQ3_S runs at up to 46.7 tok/s for local inference (best of 2 community benchmark runs across 1 GPU).

Ollamallama.cppIQ3_S
ShareRedditX

Model size

27B

Peak speed

46.7 tok/s

Average speed

46.7 tok/s

Avg PP

1358.9 tok/s

Min memory

12.1 GB

Max context

32.768 tokens

Avg output / run

42.778 tokens

Avg runtime / run

15m 30s

Avg quality

54.3

Benchmark runs

2

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 2 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)
Overall42.554.366.0
Agent Workflow80.385.390.2
Code Generation0.32.95.5
Role Play & Narrative79.784.388.8
Research & Analysis4.544.684.8

Performance by hardware and tool

Every hardware/tool/quantization combination unsloth/Qwen3.8-27B-GGUF:UD-IQ3_S has been benchmarked on, ranked by peak token generation speed. Last updated September 2026.

HardwareToolQuantBest tok/sAvg tok/sMemoryContextQualityRuns
NVIDIA GeForce RTX 4080 SUPERllama.cpp46.7 tok/s46.7 tok/s13.2 GB8.192 tokens41.21
NVIDIA GeForce RTX 4080 SUPEROllamaIQ3_S46.7 tok/s46.7 tok/s12.1 GB32.768 tokens67.31

Benchmark runs

All 2 unsloth/Qwen3.8-27B-GGUF:UD-IQ3_S runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.

Frequently asked questions

Is hf.co/unsloth/Qwen3.8-27B-GGUF:UD-IQ3_S good for coding?
In our benchmarks, hf.co/unsloth/Qwen3.8-27B-GGUF:UD-IQ3_S scores 2.9/100 for coding. It runs at about 46.7 tok/s, so if you want more speed, unsloth/Qwen3.8-27B-GGUF:IQ3_S is faster (~104.3 tok/s) and still scores well for coding (83.0/100).
Is hf.co/unsloth/Qwen3.8-27B-GGUF:UD-IQ3_S good for agentic (tool-using) tasks?
In our benchmarks, hf.co/unsloth/Qwen3.8-27B-GGUF:UD-IQ3_S scores 85.3/100 for agentic workflows. It runs at about 46.7 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 unsloth/Qwen3.8-27B-GGUF:UD-IQ3_S for local inference?
Across 2 community benchmark runs, unsloth/Qwen3.8-27B-GGUF:UD-IQ3_S reaches up to 46.7 tok/s and averages 46.7 tok/s, with the fastest results on NVIDIA GeForce RTX 4080 SUPER.
How much memory does unsloth/Qwen3.8-27B-GGUF:UD-IQ3_S need?
The leanest observed configuration used about 12.1 GB of memory (quantizations tested: IQ3_S).
Which tools have been used to run unsloth/Qwen3.8-27B-GGUF:UD-IQ3_S?
Benchmarks were submitted using Ollama, llama.cpp. Results are community-contributed and updated as new runs arrive.