DecisionBenchResults coming soon

Decision API
leaderboard

An independent, closed benchmark of Decision APIs: how often they decide right, how well their probabilities are calibrated, how consistent they are, and what they cost.

DecisionBench measures how well Decision APIs pick the right option from a closed set and how much their confidence can be trusted. Results are broken down on two levels: 3 modalities and 5 task domains.

Results coming soon. The first run on the private set has not been published yet. Scores will appear here once it is complete; nothing on this site is an estimate.
Score vs. cost
020406080100$0$1$2$3$4$5Cost per 1k decisions (USD)Score
Results coming soon
Score vs. latency
0204060801000s1s2s3s4sMedian latencyScore
Results coming soon

Results

Overall
By modality
By domain
#ModelType
—Elephant 1baseten/elephant-1Pending—————Decision API
—Joey 1baseten/joey-1Pending—————Decision API
—Mammoth 1baseten/mammoth-1Pending—————Decision API
—Layaconvaiinnovations/layaPending—————Decision API
—Gemini 3.5 Flash Litegoogle/gemini-3.5-flash-litePending—————LLM baseline
—Gemini Decision 1google/gemini-decision-1Pending—————Decision API
—D1liquid/d1Pending—————Decision API
—D1 Ultraliquid/d1-ultraPending—————Decision API
—Microsoft Decision 1microsoft/microsoft-decision-1Pending—————Decision API
—Drex Max v1.1nace-ai/drex-max-v1.1Pending—————Decision API
—Drex v2.0nace-ai/drex-v2.0Pending—————Decision API
—GPT-6 Luna Decisionsopenai/gpt-6-luna-decisionsPending—————Decision API
—Decisions Fastparallel/parallel-decisions-fastPending—————Decision API
—Decider Flash v1.2perplexity/pplx-decider-flash-v1.2Pending—————Decision API
—Decider v1.2perplexity/pplx-decider-v1.2Pending—————Decision API
—Classify 1scaledown/scaledown-classify-1Pending—————Decision API
—Jev 1.14.0typesafe-ai/jev-1.14.0Pending—————Decision API