REPRODUCIBILITY · GGUF QUANTIZATION

Quality and memory reproduced across two runs. Wall-clock latency did not.

We ran the same pinned GGUF quantization and evaluation pipeline twice — two operators, fresh containers, one pinned image — on two small Apache-2.0 models, to see which measured quantities survive a repeat.

CPU-only, no GPU. The latency failure is kept in the result, not smoothed over.

What reproduced

Swipe or use the arrow keys →

MeasurementAcross the two runsReproduced
F16 and Q4_K_M GGUF filesByte-identical (SHA-256)Yes
Perplexity, Wikitext-2 test (fixed 128 × 512-token budget)Identical to four decimals, each armYes
HellaSwag acc_norm, fixed 200-task subsetIdentical, each armYes
Peak RSS during the perplexity runWithin 0.13%Yes
Wall-clock model load timeUp to 61% spreadNo
Prompt and generation throughputUp to 34% spreadNo

Tolerance was fixed in advance at ±2% for every quantity. Both GGUF files and both evaluation sets carry their SHA-256 in the pinned inputs. The two latency rows are a property of our shared CPU host, not a comparison of the models.

Q4_K_M against F16

Swipe or use the arrow keys →

ModelPerplexity F16Perplexity Q4_K_MHellaSwag F16HellaSwag Q4_K_MPeak RSS F16Peak RSS Q4_K_M
Qwen2.5-0.5B-Instruct15.378215.784250.50%50.50%1942.6 MiB1402.0 MiB
SmolLM2-360M-Instruct14.233814.645352.50%54.50%1114.1 MiB702.8 MiB

Run 1 of 2. Perplexity is comparable only within a model, between its two arms. HellaSwag acc_norm is not expected to match between the two models; the question is whether each model’s own numbers repeat.

What this does not show

Two executions of one harness are repeat execution, not independent scientific reproduction: the same code, protocol and tolerance produced both runs. This covers two small models and one quant recipe, not GGUF quantization in general. We make no latency or throughput claim from it — reproducible wall-clock measurement needs exclusive CPU. Everything needed to rerun it is published, including the pinned image digest, model revisions and input hashes.

Download the harnessRead the two-run resultsRead the commands