run_id stringclasses 6
values | model_id stringclasses 4
values | baseline_kind stringclasses 2
values | quant_type stringclasses 4
values | baseline_bytes int64 15.2B 65.5B | quantized_bytes int64 4.68B 19.9B | compression_ratio float64 1.88 3.3 | size_reduction_fraction float64 0.47 0.7 | measurement_scope stringclasses 1
value | excludes stringclasses 1
value | baseline_runner stringclasses 2
values | quantized_runner stringclasses 2
values | execution_substrate stringclasses 2
values | baseline_tokens_per_second float64 18.7 55.4 | quantized_tokens_per_second float64 16.3 119 | observed_wall_time_ratio float64 0.85 2.7 | wall_time_direction stringclasses 2
values | throughput_ratio float64 0.87 3.63 | token_volume_cost_proxy_ratio float64 | monetary_cost_ratio_status stringclasses 1
value | artifact_footprint_status stringclasses 1
value | hardware_backend_scope stringclasses 1
value |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Mistral_Nemo_Instruct_2407_20260814_030505 | mistralai/Mistral-Nemo-Instruct-2407 | gguf_f16 | gguf_q4_k_m | 24,504,279,808 | 7,477,207,744 | 3.277197 | 0.694861 | stored_model_weight_artifact_bytes | modal_container_storage;runtime_vram | full_weight_llama_cpp | quantized_llama_cpp | local_llama_cpp | 25.043651 | 91.029623 | 2.703816 | speedup | 3.634838 | null | suppressed_missing_per_runner_cost_usd | computed | observed evaluation run on the recorded hardware and backends |
Mistral_Nemo_Instruct_2407_20260815_113254 | mistralai/Mistral-Nemo-Instruct-2407 | gguf_f16 | gguf_q5_k_m | 24,504,279,808 | 8,727,634,688 | 2.807666 | 0.643832 | stored_model_weight_artifact_bytes | modal_container_storage;runtime_vram | full_weight_llama_cpp | quantized_llama_cpp | local_llama_cpp | 30.135281 | 79.967924 | 2.564463 | speedup | 2.653631 | null | suppressed_missing_per_runner_cost_usd | computed | observed evaluation run on the recorded hardware and backends |
Mistral_Nemo_Instruct_2407_20260816_084553 | mistralai/Mistral-Nemo-Instruct-2407 | gguf_f16 | gguf_q8_0 | 24,504,279,808 | 13,022,372,608 | 1.881706 | 0.468567 | stored_model_weight_artifact_bytes | modal_container_storage;runtime_vram | full_weight_llama_cpp | quantized_llama_cpp | local_llama_cpp | 33.114068 | 58.860128 | 1.714412 | speedup | 1.777496 | null | suppressed_missing_per_runner_cost_usd | computed | observed evaluation run on the recorded hardware and backends |
Qwen2.5_14B_Instruct_1M_20260815_220633 | Qwen/Qwen2.5-14B-Instruct-1M | modal_f16 | modal_q4_k_m | 29,547,716,384 | 8,988,110,624 | 3.287422 | 0.69581 | stored_model_weight_artifact_bytes | modal_container_storage;runtime_vram | full_weight_modal_llama_cpp | quantized_modal_llama_cpp | modal_llama_cpp | 33.060657 | 31.686448 | 0.961467 | slowdown | 0.958434 | null | suppressed_missing_per_runner_cost_usd | computed | observed evaluation run on the recorded hardware and backends |
Qwen2.5_32B_Instruct_20260815_081051 | Qwen/Qwen2.5-32B-Instruct | modal_f16 | modal_q4_k_m | 65,535,969,920 | 19,851,336,320 | 3.301338 | 0.697093 | stored_model_weight_artifact_bytes | modal_container_storage;runtime_vram | full_weight_modal_llama_cpp | quantized_modal_llama_cpp | modal_llama_cpp | 18.726914 | 16.30757 | 0.853546 | slowdown | 0.870809 | null | suppressed_missing_per_runner_cost_usd | computed | observed evaluation run on the recorded hardware and backends |
Qwen2.5_7B_Instruct_20260814_234822 | Qwen/Qwen2.5-7B-Instruct | gguf_f16 | gguf_q4_k_m | 15,237,853,632 | 4,683,073,984 | 3.253814 | 0.692668 | stored_model_weight_artifact_bytes | modal_container_storage;runtime_vram | full_weight_llama_cpp | quantized_llama_cpp | local_llama_cpp | 55.4357 | 119.12482 | 1.984892 | speedup | 2.148883 | null | suppressed_missing_per_runner_cost_usd | computed | observed evaluation run on the recorded hardware and backends |
quant_eval — Efficiency and footprint
One row per published run: stored weight artifact bytes before and after quantization, compression ratio, observed evaluation wall-time ratio with an explicit direction label, and token throughput.
Part of the quant_eval public corpus: a per-case behavioral evaluation of full-weight and quantized large language models across eight agent-relevant task families, with paired statistical testing.
Cite this dataset: 10.5281/zenodo.22010723 — concept DOI, always resolves to the latest version. This exact deposit: 10.5281/zenodo.22010724 — version DOI, frozen. Cite this one where reported numbers must stay verifiable against the object referenced.
What this file contains
| File | Rows | Columns |
|---|---|---|
quant_eval_efficiency_and_footprint.csv |
6 | 22 |
Supporting files: source_bundle_checksums.json.
Corpus scope
| Run | Model | Baseline | Quantized | Substrate | Licence |
|---|---|---|---|---|---|
Mistral_Nemo_Instruct_2407_20260814_030505 |
mistralai/Mistral-Nemo-Instruct-2407 | gguf_f16 | gguf_q4_k_m | local | Apache-2.0 |
Mistral_Nemo_Instruct_2407_20260815_113254 |
mistralai/Mistral-Nemo-Instruct-2407 | gguf_f16 | gguf_q5_k_m | local | Apache-2.0 |
Mistral_Nemo_Instruct_2407_20260816_084553 |
mistralai/Mistral-Nemo-Instruct-2407 | gguf_f16 | gguf_q8_0 | local | Apache-2.0 |
Qwen2.5_14B_Instruct_1M_20260815_220633 |
Qwen/Qwen2.5-14B-Instruct-1M | modal_f16 | modal_q4_k_m | Modal | Apache-2.0 |
Qwen2.5_32B_Instruct_20260815_081051 |
Qwen/Qwen2.5-32B-Instruct | modal_f16 | modal_q4_k_m | Modal | Apache-2.0 |
Qwen2.5_7B_Instruct_20260814_234822 |
Qwen/Qwen2.5-7B-Instruct | gguf_f16 | gguf_q4_k_m | local | Apache-2.0 |
Every run evaluates a full-weight baseline and a quantized variant of the same model against the identical locked fixture set, case for case. Statistical comparison is paired: the two-sided exact McNemar test on per-case outcomes, with Wilson intervals on the rates.
Figures
Observed evaluation wall-time ratio (full weight divided by quantized) for all six published pairs. Hatched bars fall below parity — the quantized run was slower. Observed harness wall time on the recorded hardware and backends; not a controlled throughput benchmark and not a general claim about quantization performance at any precision on any hardware.
Figures are generated directly from the harness rollups by the published build tooling; no plotted value is recomputed, smoothed, or fitted.
Columns
quant_eval_efficiency_and_footprint.csv
run_id model_id baseline_kind
quant_type baseline_bytes quantized_bytes
compression_ratio size_reduction_fraction measurement_scope
excludes baseline_runner quantized_runner
execution_substrate baseline_tokens_per_second quantized_tokens_per_second
observed_wall_time_ratio wall_time_direction throughput_ratio
token_volume_cost_proxy_ratio monetary_cost_ratio_status artifact_footprint_status
hardware_backend_scope
Verification
This corpus is derived from sanitized publication bundles produced by the quant_eval harness. It is designed to be checked rather than trusted:
source_bundle_checksums.json, included here, republishes, verbatim, the SHA-256 digest and byte length of every file in every source bundle. No source file was modified.- Before this file was written, the builder verified all 72 source-file digests and independently recomputed all 96 family x runner pass rates from the raw per-case rows, matching the harness rollups exactly.
- The figures here are not derived from the per-case results dataset. Stored-artifact byte counts and observed wall time are recorded by the harness at run time and are carried through from the source bundles unchanged; they are traceable through the bundle digests above, not recomputable from per-case rows. Pass-rate aggregates, which are recomputable, live in the paired degradation statistics (D5) and family pass rates (D6) datasets.
Limits you should know before using this
- Decoding conditions are not uniform across models. Temperature follows each publisher's own model card, so cross-model comparison of absolute pass rates is confounded. Within-run pairing is unaffected, which is what the paired test requires. The conditions are published per row and per run so they can be filtered on.
- Runs on the Modal substrate record
seedstatusunsupported— the deployed method signature accepts no seed — so those runs are not exactly reproducible. Local runs applied a fixed seed. - Runtime figures are observed harness wall time on the recorded hardware and backends. They are not a controlled throughput benchmark and not a general claim about quantization performance at any precision on any hardware. Direction is published as an explicit label because not every measured pair is a speedup.
- The fuzz family is an adaptive trajectory evaluated from identical starting fixtures. Its paired test compares complete case outcomes, not identical post-divergence prompts.
- Calibration runs are not published. Runs that informed a published run are disclosed by identifier in
calibration_lineage.csv, published in the run provenance dataset (D4), so the record is complete without releasing provisional numbers.
Citation
@dataset{pbh_quant_eval_d7,
author = {Hill, Patrick},
title = {quant_eval Efficiency and footprint},
publisher = {PBH Applied Systems, LLC},
year = {2026},
doi = {10.5281/zenodo.22010723},
note = {Version DOI: 10.5281/zenodo.22010724},
license = {CC-BY-4.0}
}
Licence
Creative Commons Attribution 4.0 International (CC BY 4.0). See LICENSE. Commercial use is permitted; attribution is required.
This corpus describes third-party models and redistributes no model weights. Each evaluated model remains under its own licence, recorded per run in the run provenance dataset.
Produced by build_datasets.py 2.4.0 from quant_eval publication bundles. Built 2026-08-19.
- Downloads last month
- 36
