Model benchmarks

Qwen3.8-9B-Distill-uncensored-heretic.i1-IQ3_M local LLM performance

As of September 2026, Qwen3.8-9B-Distill-uncensored-heretic.i1-IQ3_M runs at up to 34.3 tok/s for local inference (best of 3 community benchmark runs across 1 GPU).

llama.cpp
ShareRedditX

Model size

9B

Peak speed

34.3 tok/s

Average speed

34.3 tok/s

Avg PP

295.1 tok/s

Min memory

4.4 GB

Max context

65.536 tokens

Avg output / run

7.060 tokens

Avg runtime / run

3m 36s

Avg quality

59.7

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)
Overall52.859.764.6
Agent Workflow64.468.572.5
Code Generation25.236.347.5
Role Play & Narrative53.369.685.0
Research & Analysis63.364.566.3

Performance by hardware and tool

Every hardware/tool/quantization combination Qwen3.8-9B-Distill-uncensored-heretic.i1-IQ3_M has been benchmarked on, ranked by peak token generation speed. Last updated September 2026.

HardwareToolQuantBest tok/sAvg tok/sMemoryContextQualityRuns
Tesla P100-PCIE-16GBllama.cpp34.3 tok/s34.3 tok/s4.4 GB65.536 tokens59.73

Benchmark runs

All 3 Qwen3.8-9B-Distill-uncensored-heretic.i1-IQ3_M runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.

Frequently asked questions

Is Qwen3.8-9B-Distill-uncensored-heretic.i1-IQ3_M good for coding?
In our benchmarks, Qwen3.8-9B-Distill-uncensored-heretic.i1-IQ3_M scores 36.3/100 for coding. It runs at about 34.3 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 Qwen3.8-9B-Distill-uncensored-heretic.i1-IQ3_M good for agentic (tool-using) tasks?
In our benchmarks, Qwen3.8-9B-Distill-uncensored-heretic.i1-IQ3_M scores 68.5/100 for agentic workflows. It runs at about 34.3 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 Qwen3.8-9B-Distill-uncensored-heretic.i1-IQ3_M for local inference?
Across 3 community benchmark runs, Qwen3.8-9B-Distill-uncensored-heretic.i1-IQ3_M reaches up to 34.3 tok/s and averages 34.3 tok/s, with the fastest results on Tesla P100-PCIE-16GB.
How much memory does Qwen3.8-9B-Distill-uncensored-heretic.i1-IQ3_M need?
The leanest observed configuration used about 4.4 GB of memory.
Which tools have been used to run Qwen3.8-9B-Distill-uncensored-heretic.i1-IQ3_M?
Benchmarks were submitted using llama.cpp. Results are community-contributed and updated as new runs arrive.