ASI Alliance's Super Decider Ranks First Among 280 Systems on JevBench With 78.5 Score

Super Decider, a free AI model from Singularity Compute, now ranks first on JevBench, a public leaderboard that scores AI decision models. Its composite score of 78.5 placed it ahead of 279 other systems on the board. The release was flagged by @ASI_Alliance, the group behind $FET, and the model is free to use through ASI:Cloud (@CUDOS_) while supply lasts.

What Super Decider Does

Super Decider is built to select answers rather than write text. According to a post from Singularity Compute shared on LinkedIn, it returns choices, yes/no answers and scores, each with calibrated probabilities, and is aimed at the routine decisions made by AI agents and automated systems. In other words, the model reports how confident it is in each answer, not just the answer itself.

JevBench's published leaderboard lists the model's composite score of 78.5 at number one among 280 ranked systems, priced at $0.016. The same page labels it an open-weights model and breaks the composite into four axes: intelligence, calibration, speed and cost. Calibration measures how closely a model's stated confidence matches how often it is actually right.

The same leaderboard lists a variant, Super Decider (thinking), at 74.5 with a higher cost of $0.028. The Singularity Compute post cites a median latency of 0.25 seconds and a price of $0.016 per 1,000 decisions for the standard model.

ASI Alliance and ASI:Cloud

The model is distributed through ASI:Cloud, the alliance's compute platform. According to the ASI Alliance's launch announcement, SingularityNET provides the inference layer and model catalog, while CUDOS supplies the cloud infrastructure across 13 regions. The platform entered beta in September 2025.

The ASI Alliance describes itself as an open-source initiative focused on decentralized artificial general intelligence, formed in April 2024. It currently unites SingularityNET, Fetch.ai and CUDOS. Its governance page says activity now runs through the $FET token.

JevBench is run by an independent maintainer through the Benchmark Heaven project. Results on it reflect the benchmark's own methodology and item set, so a first-place ranking applies to that test and does not necessarily carry over to other evaluations.

Sources

JevBench: all decision models leaderboard
ASI Alliance: ASI:Cloud launch announcement
ASI Alliance: About