Kahn1 4B (Okura66, Qwen3.5-4B LoRA merge, author's sysone engine)
Data: JevBench v1.6.1, published 2026-10-06
Capability 61.7 · rank #34 of 92 on the open-weights board
Last measured: 2026-10-07 · measurement release: v1.6.1
apache-2.0
intelligence
39.7
calibration
83.8
speed
90.0
cost
46.4
Composite (secondary)
31.1 · Composite rank #44
Cost and measurement conditions
$0.061 per 1,000 decisions (estimated)
ESTIMATE, ours (the author's server returns no token usage; Florian 7 Oct 2026: use our own estimate): 829,523 input tokens counted with the author's tokenizer and prompt builder over the 1,624 v1.5 items x n_permutations 3 (the author's server default) = 2,488,569 tokens / 1,624 decisions x USD 0.04 per 1M input (deepinfra:Qwen/Qwen3.5-4B, frozen 25 Sep snapshot), 0 output tokens = USD 0.061295 per 1,000 decisions; carried like every Qwen3.5-4B row on this board. Cross-check: the same count on the v1.6 common cost basis (1,477 items, x3) gives USD 0.0597 per 1,000 (within 3 %)
Measured p50 latency: 0.066 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: 0.28 s (est. incl. network (2x + 150 ms))
Self-hosted model: measured p50 0.066 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 = 0.28 s. Model authors can ask us to time a hosted endpoint instead (on request).
x2 + 0.15 s (assumption, not measured)
evaluator-owned Lium GPU pod, 1x NVIDIA RTX 5090, the author's own sysone engine (github.com/Okura66/kahn1) on vLLM 0.30.0 with the author's documented max-model-len 4096 and calibration.json, offline (all credentials unset), driven over loopback only by the JevBench client through a thin pre-registered adapter for the author's /v1/evaluate/jev endpoint; label-free inputs; pod destroyed afterwards; JevBench add-request, measured 7 Oct 2026