Quyet-1.2-Largev1.7 · 10 s rule
Data: JevBench v1.8.15, published 2026-11-17 · method v1.7 (10 s rule), measured under the rule
Capability 81.7 · Not ranked in the current release
Added after the v1.8.10 ranking review; it is ranked at the next review.
Last measured: 2026-11-16 · measurement release: v1.8.14
Apache-2.0
intelligence
74.5
calibration
89.0
speed
88.3
cost
61.7
Composite (secondary)
76.6 · Not ranked on this board
Cost and measurement conditions
$0.019 per 1,000 decisions (estimated)
GPU-time estimate from v1.8.0: one monthly-rented RTX PRO 6000 96 GB, USD 1.5235/hour; average load 70%. Architecture: dense gemma, 31B active parameters; throughput 10106.516129 tokens/s (architecture-class calibration, efficient-engine prefill). GPU-time cost USD 0.018859030495733/1,000; old list-price estimate USD 0.0283740564137005/1,000 (google/gemma-4-31b-it at USD 0.09/M input and USD 0.34/M output). Applied min(old list-price estimate, GPU-time cost) = USD 0.018859030495733/1,000. Estimated, not a bill.
Measured p50 latency: 0.079 s (on-GPU, no network, loopback; not directly comparable with API latency, which we time end to end from our measuring host, network included)
Ranking and cap use: 0.31 s (est. incl. network (2x + 150 ms))
Self-hosted model: measured p50 0.079 s on-GPU, no network (timed on our GPU over loopback). Not directly comparable with API latency, which we time end to end from our measuring host, network included. The ranking and latency cap use the same estimate for every self-hosted row: 2 × measured + 150 ms = 0.31 s. Model authors can ask us to time a hosted endpoint instead (on request).
x2 + 0.15 s (assumption, not measured)
evaluator-owned Verda GPU VM (RTXPRO6000), offline read-only container (--network none); the author's quyet 1.2.0 package (github.com/ncchinh/quyet @d41fa079) loaded in process by the fixed pod driver (quyet.load(model_dir, device='cuda').predict); merged BF16 Gemma 4 31B + F32 option head, SDPA, one forward pass per question, no generation, no prefix cache; inputs over the 32,768-token complete-input limit or over 256 options are refused (QuestionError, scored as 422/wrong); official PyTorch 2.14.0 cu130 image + pinned overlay (transformers 5.17.0, accelerate 1.15.0, Python 3.12)
Measured latency over all decisions: p50 0.079 s · p95 0.166 s
— decisions over 10 s.
A decision that takes longer than 10 seconds (measured time) counts as a wrong answer and as zero calibration for that item: each type's calibration is multiplied by the share of its decisions answered within 10 s. The limit uses the measured time (raw). The displayed adjusted time for self-hosted rows (2 × + 0.15 s) is an assumption and is not used for the limit.
Model and serving disclosure
Paid public evaluation. Measured on the 27 Oct draw on an evaluator-owned Verda GPU VM (RTXPRO6000), offline with the author's serving stack at the pinned commit. Failed or unanswered requests count as wrong answers. Successor of Quyet-1.0-Large (measured on the v1.5 draw); no comparison is drawn between the two draws. Decision model: Gemma 4 31B with fully merged weights and a separate option head; choice via the option head, score and noul via vocabulary-symbol scores with fixed per-type temperatures; 0 output tokens.