Microsoft-Decision-1 (updated weights, Oct 2026)v1.7 · 10 s rule
Data: JevBench v1.7.15, published 2026-10-28 · method v1.7 (10 s rule), measured under the rule
Capability 79.8 · API offering, listed but not ranked on the API leaderboard
Measured on the 27 Oct 2026 draw, which is not yet equated to the reference scale: its first anchor runs used models weaker than this row, and stronger anchors are being measured. Shown unranked until then.
Last measured: 2026-10-28 · measurement release: v1.7.15
Licence not stated
intelligence
67.1
calibration
92.5
speed
89.3
cost
66.4
Composite (secondary)
77.0 · Not ranked on this board
Cost and measurement conditions
$0.013 per 1,000 decisions (not published)
Advertised Azure Foundry tariff USD 0.042 per million input tokens, output free (same basis as the first Microsoft-Decision-1 row); provider-reported usage on all n = 1,493 decisions (467,021 input and 1,489 output tokens), the 4 refused requests kept in the denominator: USD 0.013138 per 1,000 decisions. Billed invoice unknown.
p50 latency: 0.3 s
none (API)
Vendor-hosted Azure AI Foundry deployment (Sweden Central), native API; latency measured from our server in Germany. 1,489 of 1,493 requests answered (HTTP 200); 4 were refused by the server because the prompt exceeded its 32,768-token context limit and are scored wrong; no decision took longer than 10 s.
Measured latency over 1,493 decisions: p50 0.319 s · p95 0.364 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.