JevBench by Benchmark Heaven · released v1.6.1
How to choose a Jev-class model
Start with the use case, then read the metric behind each recommendation. The guide uses the latest published JevBench artifact; it does not treat the overall composite as raw accuracy.
Data: JevBench v1.6.1, published 2026-10-06
Most accurate (Intelligence axis)
Quyet-1.0-Large (Chinh Nguyen, Gemma-4-31B decoder, option-letter logits)
73.4
Published Intelligence axis in v1.6.1. The JevBench Score is a separate composite.
Fastest (Speed axis)
Quyet-1.0-Tiny (Chinh Nguyen, mmBERT-small encoder (16 layers kept) with fixed heads)
95.1of 100
This is the published Speed axis. Use the row’s latency conditions when estimating real performance; it is not a universal wall-clock guarantee.
Cheapest per decision (Cost axis)
verdict-small (Manavarya09, multilingual-e5-small 118M)
100.0of 100
Published cost: $0.0009 per 1,000 decisions (estimated)
Cost axis is a comparison score. The row’s cost basis distinguishes measured, estimated and announced values.
Self-hosting evidence
Use rows with public openness, license and repository evidence
The list below is ordered by the published JevBench rank. The openness status and license terms are copied from each public aggregate row.
Systems with explicit self-hosting evidence
Quyet-1.0-Large (Chinh Nguyen, Gemma-4-31B decoder, option-letter logits) · open-weights rank 1
Weights marked open in the published row · apache-2.0
René-1 31B FP8 (salfatigroup, Gemma 4 31B full fine-tune, one-pass option readout) · open-weights rank 4
Weights marked open in the published row · apache-2.0
H2O-Lightning-4B v1.1 (H2O.ai, Qwen3.5-4B fine-tune, stock vLLM + open shim) · open-weights rank 5
Weights marked open in the published row · apache-2.0
Bobcat Flash 1.2 (Gemma-4-26B-A4B-it + two merged rank-64 LoRA adapters, typed-decision readout at the first answer position, FP8, self-hosted) · open-weights rank 6
Weights marked open in the published row · apache-2.0
Surogate Rune 26B-A4B v3 (v1.6 pool) · open-weights rank 7
Weights marked open in the published row · apache-2.0
decider-12b v2 (Mapika) · open-weights rank 8
Weights marked open in the published row · apache-2.0
Xor 26B-A4B (Juspay, Gemma-4-26B-A4B, bf16) · open-weights rank 9
Weights marked open in the published row · apache-2.0
decider-12b v1, stock Gemma-4-12B-it (Mapika) · open-weights rank 10
Weights marked open in the published row · apache-2.0
Rows without explicit openness, license or repository evidence are left unclassified. Confirm each model’s terms and requirements before using it in production.
Keep the measures separate
Arithmetic mean of Intelligence and Calibration; ranked models within both official caps Accuracy, speed, cost and the composite answer different questions. Review the full live board and its row disclosures before making a deployment choice.
Frequently asked questions
- Which Jev-class model has the highest published Intelligence?
- Quyet-1.0-Large (Chinh Nguyen, Gemma-4-31B decoder, option-letter logits) has the highest Intelligence axis among the open-weights ranked rows of v1.6.1 (73.4). This is a separate measure from the composite JevBench Score.
- Which Jev-class model is fastest?
- Quyet-1.0-Tiny (Chinh Nguyen, mmBERT-small encoder (16 layers kept) with fixed heads) has the highest published Speed axis score (95.1). The axis is a benchmark score; it does not promise the same wall-clock latency on every host or workload.
- Which Jev-class model is cheapest?
- verdict-small (Manavarya09, multilingual-e5-small 118M) has the highest Cost axis score (100.0). Its published cost is estimated at $0.0009 per 1,000 decisions; check the board’s row disclosure for the basis before comparing bills.
- How do I find a self-hostable Jev alternative?
- Use only systems marked open or open weights in the published row, with a license note and public repository link. This page leaves absent openness, license or repository evidence unclassified; review the linked terms before deployment.