Wald 4B v2.1
Data: JevBench v1.6.39, published 2026-10-22
Capability 69.3 · rank #28 of 139 on the open-weights board
Last measured: 2026-10-10 · measurement release: v1.6.5
Apache-2.0 (weights and server code)
intelligence
52.0
calibration
86.6
speed
93.1
cost
46.3
Composite (secondary)
54.3 · Composite rank #37
Cost and measurement conditions
$0.062 per 1,000 decisions (estimated)
Self-hosted open weights, no bookable tariff. v1.5-M2 base-model reference rule: base model Qwen/Qwen3.5-4B (model card base_model) -> DeepInfra Qwen/Qwen3.5-4B, USD 0.03/M input and USD 0.15/M output, frozen 25 Sep 2026 price cut-off snapshot (METHOD-v1.5-ADDENDUM-PRICING-INTERPRETATION-1 I-2/I-3; listing deprecated after the cut-off, stays the frozen v1.5 reference). Same basis as the published Wald 4B (v2.0) row. Read 2026-10-10.
Measured p50 latency: 0.027 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: 0.20 s (est. incl. network (2x + 150 ms))
Self-hosted model: measured p50 0.027 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 = 0.20 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 (RTX PRO 6000 96GB), offline read-only container (--network none); official vllm/vllm-openai:v0.30.0@sha256:8a69ffad... image; author's server wald_serve.native_vision from the model repo, --prompt-format plain (author RUNBOOK's JevBench format), serving.json defaults otherwise (effort none = one pass, noul_decisive_floor 0.81)