VisionJEV 1.0.0v1.7 · 10 s rule
Data: JevBench v1.7.23, published 2026-10-29 · method v1.7 (10 s rule), measured under the rule
Capability 86.5 · Not ranked in the current release
Measured on the 27 Oct 2026 draw, which is not yet equated to the reference scale; stronger anchor runs for that draw are being measured. Shown unranked until then.
Last measured: 2026-10-29 · measurement release: v1.7.21
Apache-2.0 + NOTICE
intelligence
78.8
calibration
94.3
speed
84.8
cost
34.1
Composite (secondary)
28.8 · Not ranked on this board
Cost and measurement conditions
$0.158 per 1,000 decisions (estimated)
ESTIMATE (*est.), frozen base-model reference Qwen/Qwen3.8-27B at USD 0.42/M input and 3.00/M output; not an author tariff or pod price; output 0 under the zero-output convention. The frozen pricing_v15 module (sha256 645fd82eac728d597048bec0995ecd3ac896566e269fb6c9a1fdb84c171a9ac4, re-frozen 26 Oct 2026) lists the card's base string Qwen/Qwen3.8-27B and prices it as qwen/qwen3.8-27b (USD 0.42/M input, USD 3.00/M output) (checked on this module). No explicit reference or rate override is used. Estimated, not charged, not a quote from the author or any host. INPUT BASIS: the usage block reports the observed input-id length of each answered request: 562,151 input tokens across the 1,495 answered items, identical to our first-party offline count of the same prompts. The 5 items refused with HTTP 422 at the model's 8,192-token input cap carry NO usage, are never billed, and stay in the denominator of the per-1,000-decisions cost. OUTPUT: the model is a classifier and generates no text; output_tokens is 0 and, under the zero-output convention for native-probability rows, output is priced at USD 0/M. Native usage on the fixed 1,489-item basis: 559,606 input tokens and 0 output tokens on the 1,489 common items. COST FIGURES: USD 0.157847 per 1,000 decisions, 2.44x the official cost cap USD 0.064595. Cost on the 1,489 items every row on this draw answered (same-items rule); all-items cost USD 0.157402 per 1,000 decisions.
Measured p50 latency: 0.152 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: 0.45 s (est. incl. network (2x + 150 ms))
Self-hosted model: measured p50 0.152 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 = 0.45 s. Model authors can ask us to time a hosted endpoint instead (on request).
x2 + 0.15 s (assumption, not measured)
Self-hosted on one evaluator-rented Verda NVIDIA A100-SXM4-80GB in Finland (EU), using the author's own documented HTTP server on loopback; BF16 and SDPA. Measured 29 Oct 2026. Public warm-up preceded timed measurements. The model, source and input were read-only, network cut before scored input. GPU rental is separate from the estimated model tariff.
Measured latency over 1,500 decisions: p50 0.152 s · p95 0.290 s
0 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.