# QEV Compact decision model by Ken Jo

> QEV Compact prunes two multimodal models to 466M and 1.31B parameters. Read its open weights, public vision results and calibration limits.

Updated 2026-11-12 (Australia/Adelaide)

## Quick answers

### What is QEV Compact by Ken Jo?

[QEV Compact](<https://decisionmodels.ai/models/qev-compact>): Two pruned QEV v3 models for text and image questions.

[QEV Compact facts](<https://decisionmodels.ai/models/qev-compact>)

[Source](<https://huggingface.co/ken-jo/QEV-2B-compact>)

### Who makes QEV Compact?

[QEV Compact](<https://decisionmodels.ai/models/qev-compact>): Ken Jo

[QEV Compact facts](<https://decisionmodels.ai/models/qev-compact>)

[Source](<https://huggingface.co/ken-jo/QEV-2B-compact>)

By Ken Jo · Hugging Face weight release: 2026-11-09 · OPEN

Two pruned QEV v3 models for text and image questions. The maker removes eight of 24 language-model blocks and trims the input vocabulary to 32,768 rows, then distils a decision readout. They return option probabilities, a typed answer and a separate abstain flag. Ken Jo's QEV family is separate from labqoat's [Qev 0.8B](<https://decisionmodels.ai/models/qev>).

## Facts

- Sizes: 466,313,408 and 1,314,213,824 parameters, per the cards
- Inputs: English text and up to four images; Choice, Noul and Score
- Context: 8,192 input tokens; over-budget requests refused
- Runtime: Bundled backbone and head; pinned qev Python loader on CUDA, accurate mode only
- Licence: Apache-2.0 files included; training datasets retain their own terms

## Claims and sources

- **VENDOR-REPORTED**: The maker reports public vision index scores of 52.52 and 62.15 on 12,210 rows across 11 benchmarks, versus 52.70 and 62.27 for the corresponding full-depth parents. These are public-only normalized scores, not accuracy percentages; private-half results are pending. [source](<https://huggingface.co/ken-jo/QEV-2B-compact>)
- **VENDOR-REPORTED**: On an RTX 4060 Ti 8 GB at batch 1, the maker reports peak GPU memory of 1.15 GiB and 2.76 GiB, with median p50 timings of 54.4 ms and 99.0 ms across 47 request profiles. Calibration and abstention thresholds were fitted on development data; the empirical error target is not a guarantee on new data. [source](<https://huggingface.co/ken-jo/QEV-0.8B-compact>)

## Links

- [Source: 2B compact card and weights](<https://huggingface.co/ken-jo/QEV-2B-compact>)
- [Source: 0.8B compact card and weights](<https://huggingface.co/ken-jo/QEV-0.8B-compact>)
- [Source: Licence](<https://huggingface.co/ken-jo/QEV-2B-compact/blob/main/LICENSE>)

## Related projects

No related projects are listed in the registry.