Enabling new use cases for decision models beyond high-volume cheap decisions

Expected valueAverage rank on 6 benchmarksUpright Eevo-1: average rank 1.0Perplexity Decider-1: average rank 2.4TypeSafe Jev 1.14: average rank 3.1Cloudflare Clef 27B: average rank 4.7OpenAI Decisions: average rank 4.8Microsoft Decision-1: average rank 5.0Economic rationalitySTEER Public eval81.376.671.768.864.866.4Bayesian reasoningCogBench (urn) Public eval98.197.797.690.383.982.7Strategic reasoning with incomplete informationPokerBench (pre-flop) Public eval87.777.374.370.281.076.3Game-theoretic reasoningPayoff games Upright eval93.381.281.275.871.572.1Risk and sequential decisionsBackgammon doubling cube Upright eval82.169.374.762.168.155.3Expected-value reasoningEV decisions Upright eval84.181.680.066.665.967.2Reasoning & ClassificationAverage rank on 23 benchmarksUpright Eevo-1: average rank 1.2Perplexity Decider-1: average rank 2.6TypeSafe Jev 1.14: average rank 4.0Cloudflare Clef 27B: average rank 3.3OpenAI Decisions: average rank 5.7Microsoft Decision-1: average rank 4.2Hard multi-step reasoningBBH Public eval94.778.690.161.851.364.6Reasoning over long narrativesMuSR Public eval84.653.556.574.230.734.7Causal reasoningCLadder Public eval90.681.677.288.224.050.6Arithmetic word problemsGSM8K Public eval85.480.567.852.123.257.7Reasoning about codeCRUXEval Public eval91.977.759.978.646.046.6Commonsense reasoningWinoGrande Public eval91.388.082.287.147.979.2Chess move choiceChessBench Public eval25.717.99.517.810.011.3Select-all-that-apply questionsSATA-Bench Public eval60.922.126.932.630.235.0Contract inferenceContractNLI Public eval80.176.957.773.066.966.4Adversarial inferenceANLI Public eval73.663.361.555.051.364.5Commonsense completionHellaSwag Public eval98.095.592.597.687.992.8Aspect-based sentimentACOS Public eval66.439.131.031.011.719.8Financial sentimentFinEntity Public eval92.791.781.994.373.887.8Sarcasm detectioniSarcasmEval Public eval56.467.439.931.019.149.9Stance detectionVAST Public eval65.170.360.539.036.053.7Clinical-trial inferenceNLI4CT Public eval75.072.371.166.956.667.6Hallucination detectionRAGTruth Public eval70.267.748.457.641.044.1Intent classificationBANKING77 Public eval92.590.979.794.076.386.7Intent and out-of-scope detectionCLINC150 Public eval97.694.490.297.488.886.3Reasoning-intensive retrievalBRIGHT Public eval41.643.640.140.632.740.1Product search relevanceAmazon ESCI Public eval53.952.946.346.543.249.4Phishing detectionPhishNChips Public eval94.449.367.558.73.831.8Multi-hop claim verificationHoVer Public eval81.272.553.531.241.950.4Tools & AutomationAverage rank on 5 benchmarksUpright Eevo-1: average rank 1.6Perplexity Decider-1: average rank 2.6TypeSafe Jev 1.14: average rank 4.7Cloudflare Clef 27B: average rank 2.2OpenAI Decisions: average rank 6.0Microsoft Decision-1: average rank 3.9Function callingBFCL Public eval97.897.095.197.992.295.5Tool retrievalToolRet Public eval64.162.459.664.353.060.5Next API callAPI-Bank Public eval88.887.286.491.885.086.4Device controlHome appliance simulator Public eval100.090.945.580.742.150.0When to call a toolWhen2Call Public eval88.687.175.563.360.782.3Arts & TasteAverage rank on 6 benchmarksUpright Eevo-1: average rank 1.3Perplexity Decider-1: average rank 2.3TypeSafe Jev 1.14: average rank 4.5Cloudflare Clef 27B: average rank 3.1OpenAI Decisions: average rank 5.3Microsoft Decision-1: average rank 4.4Poetic formBPoMP Public eval98.090.580.894.181.386.9HumourHumicroedit Public eval37.831.623.533.122.824.1Chord recognitionPOP909 Public eval76.960.215.314.817.141.9Colour preferencecfcolor Public eval36.041.130.134.021.824.7Group preference predictionHabermas Machine Public eval46.840.319.654.517.914.2Caption matchingNew Yorker Public eval83.773.765.762.860.762.8Knowledge & TriviaAverage rank on 3 benchmarksUpright Eevo-1: average rank 2.0Perplexity Decider-1: average rank 3.5TypeSafe Jev 1.14: average rank 1.0Cloudflare Clef 27B: average rank 4.5OpenAI Decisions: average rank 4.8Microsoft Decision-1: average rank 5.2Graduate-level scienceGPQA Diamond Public eval46.340.167.431.334.028.6Academic knowledgeMMLU-Pro Public eval71.568.479.961.752.460.4Expert-level questionsHLE Public eval2.10.03.50.00.00.0

Sources: Jev Decision Index (board, 22 Oct 2026) · our evaluation runs (24–26 Oct 2026) · OpenAI Decisions and Microsoft Decision-1 on the Decision Index: our runs of each vendor’s decision API (24 Oct 2026) · STEER (Raman et al. 2024, CC BY 4.0) · CogBench (Coda-Forno et al., ICML 2024) · Habermas Machine (Google DeepMind, CC BY 4.0).

Average rank = mean rank per category, among models with a result; ties averaged.

Compared to trillion-parameter generative frontier models, decision models are relatively easy and cheap to train and operate. This makes it possible for organisations to have control and sovereignty over their critical AI infrastructure, running models like Eevo on-premise, or even creating their own decision models optimized for their use, having full control over the whole stack.

At Upright, we are flexible on how organisations use Eevo. Organisations can choose whether to use Eevo's public cloud version (eevo.io) or deploy it on-premise (managed by us or the organization). For advanced use cases, we also create custom versions of Eevo optimized for your use case, with the same flexible deployment options.

For more information on Eevo, read the announcement, or try it yourself.

For more information, contact juho.ojala@uprightproject.com.