ไธญๆ–‡
โ† All articles

RESEARCH / STRUCTURED DECISIONS

SPX-CD-PreviewOpen-source multimodal fast decision models.

SPX-CD-Flash โ†— ยท SPX-CD-Pro โ†— ยท SPX-CD-Omni โ†—

SPX-CD (Simplex Calibrated Decision Model) uses text, images, and candidate options to quickly return choices, probabilities, and scores your application can use.

API calls โ†—Run test cases โ†—

SPX-CD / IN ACTION

01 / 07
SHOWCASEScroll for more โ†“
Photo credits โ†—

02 / LIVE BROWSER

Give it a task. Watch it work.

Try flight planning, product search, or form filling.

200 API calls

Each inference uses one API call, up to 25 calls per task. IKEA visits a live website; other tabs use interactive sample pages. No purchases or external form submissions.

โ— โ— โ—Simplex BrowserThis task: 0 / 25 API callsReady
Live browser screenshot
Enter a task to watch the browser work and see the result.

    03 / TEST CASES

    Try these decision tasks.

    Choose an example to try multiple questions, multi-select, scoring, or image-based decisions.

    Edit request JSON

    Model results

    Run an example to see the modelโ€™s answers.

    04 / EVALUATION

    Evaluation results

    CMDB-1500, JevBench and Decision Index compare decision accuracy and probability calibration.

    CMDB-1500

    1,200 text tasks and 300 image tasks. SPX-CD uses effort=2 in this evaluation.

    SPX-CD-Pro81.20%Overall accuracy ยท 1,500 questions
    SPX-CD-Flash79.53%Overall accuracy ยท 1,500 questions
    SPX-CD-Omni76.27%Overall accuracy ยท 1,500 questions
    Filter
    SPX-CDJevGeneral reasoning modelsOther decision models

    View exact text results
    CMDB-1500 Text results; fixed total 1,200, valid answers counted over the full evaluated scope
    RankModel / versionAccuracy ยท 0โ€“100%Correct / fixed totalValid / evaluated scope
    01GPT 6.1 Sol ยท low80.00%960 / 1,2001,500 / 1,500
    02SPX-CD-Pro79.33%952 / 1,2001,500 / 1,500
    03Claude Sonnet 5.5 ยท low78.00%936 / 1,2001,500 / 1,500
    04StartLux-Decision-27B ยท BF1677.92%935 / 1,2001,500 / 1,500
    05pplx-decider ยท v1.1 27B77.83%934 / 1,2001,500 / 1,500
    06SPX-CD-Flash77.58%931 / 1,2001,500 / 1,500
    07GPT 6 Luna ยท low76.58%919 / 1,2001,500 / 1,500
    08Microsoft-Decision-1 ยท9B76.17%914 / 1,2001,200 / 1,200
    09GLM 5.3 Flash ยท low75.75%909 / 1,2001,500 / 1,500
    10StartLux-Decision-35B-A3B ยท BF1675.67%908 / 1,2001,500 / 1,500
    11SPX-CD-Omni74.67%896 / 1,2001,500 / 1,500
    12Mercury Decide73.92%887 / 1,2001,200 / 1,200
    13pplx-decider 27B72.83%874 / 1,2001,500 / 1,500
    14Jev72.17%866 / 1,2001,200 / 1,200
    15Open-Jev ยท Zefan 27B71.67%860 / 1,2001,199 / 1,200
    16Cloudflare Clef71.50%858 / 1,2001,485 / 1,500
    17JPT ยท 9B71.33%856 / 1,2001,500 / 1,500
    18Qwen 3.8 Flash71.08%853 / 1,2001,500 / 1,500
    19Kimi K2.671.00%852 / 1,2001,500 / 1,500
    20Mapika ยท Decider 4B v2.169.83%838 / 1,2001,200 / 1,200
    21Cloudflare Clef-flash69.67%836 / 1,2001,485 / 1,500
    22GPT 6 Luna Decisions69.67%836 / 1,2001,498 / 1,500
    23Imajev-9B68.83%826 / 1,2001,478 / 1,500
    24Kev ยท 9B68.67%824 / 1,2001,200 / 1,200
    25DeepSeek V4.1 Flash68.58%823 / 1,2001,500 / 1,500
    26JPT ยท 4B67.67%812 / 1,2001,500 / 1,500
    27Decision Lux ยท 9B67.42%809 / 1,2001,200 / 1,200
    28Quyet-1.0 ยท Large 31B67.08%805 / 1,2001,085 / 1,200
    29Jet ยท v6.2 4B66.92%803 / 1,2001,200 / 1,200
    30Open-Jev ยท Zefan 9B66.67%800 / 1,2001,199 / 1,200
    31Wity-1 ยท auto66.67%800 / 1,2001,426 / 1,500
    32Nimble ยท v2 9B66.58%799 / 1,2001,200 / 1,200
    33Hopper G ยท 4B v1.365.92%791 / 1,2001,200 / 1,200
    34Malkuth ยท 4B65.58%787 / 1,2001,200 / 1,200
    35Hopper ยท 4B65.50%786 / 1,2001,200 / 1,200
    36Winnow ยท 12B Q865.33%784 / 1,2001,410 / 1,500
    37JevK564.83%778 / 1,2001,200 / 1,200
    38Lev ยท 4B64.58%775 / 1,2001,200 / 1,200
    39Kev ยท4B API64.33%772 / 1,2001,200 / 1,200
    40Mica ยท 4B BF16 GGUF64.25%771 / 1,2001,200 / 1,200
    41Kev ยท 4B64.17%770 / 1,2001,200 / 1,200
    42Jev-Omni ยท 12B63.58%763 / 1,2001,493 / 1,500
    43APUS OpenJev ยท 9B 3000 High61.42%737 / 1,2001,085 / 1,200
    44OpenDecider ยท Small61.25%735 / 1,2001,200 / 1,200
    45APUS OpenJev ยท 9B 5949 High61.08%733 / 1,2001,085 / 1,200
    46OpenDecider ยท Small TD60.67%728 / 1,2001,200 / 1,200
    47APUS OpenJev ยท 4B 5949 High60.08%721 / 1,2001,085 / 1,200
    48Ateve Jev59.75%717 / 1,2001,087 / 1,200
    49Bocha Jev58.92%707 / 1,2001,087 / 1,200
    50Winnow ยท E4B Q858.92%707 / 1,2001,410 / 1,500
    51Reflex ยท Stable 4B58.50%702 / 1,2001,387 / 1,500
    52Plumb ยท v5.2 4B58.08%697 / 1,2001,085 / 1,200
    53FLock THIS/THAT57.67%692 / 1,2001,156 / 1,200
    54Kev ยท 0.8B51.75%621 / 1,2001,200 / 1,200
    55Bosun v3.1 ยท 1.7B49.75%597 / 1,2001,200 / 1,200
    56OpenDecider ยท Nano46.75%561 / 1,2001,143 / 1,200
    57CalDec ยท Laya46.08%553 / 1,2001,189 / 1,200
    58Bosun v3.1 ยท 0.6B45.58%547 / 1,2001,200 / 1,200
    59Laya ยท Typed Decisions44.58%535 / 1,2001,189 / 1,200
    60CalDec ยท GLiNER43.17%518 / 1,2001,200 / 1,200
    61Laya ยท English41.00%492 / 1,2001,189 / 1,200
    62Laya ยท Multilingual40.25%483 / 1,2001,192 / 1,200
    63GLiNER2.5-Decide38.58%463 / 1,2001,200 / 1,200
    64Julia-135.75%429 / 1,2001,079 / 1,200
    65Liquid AI ยท d1-omni-600M35.50%426 / 1,2001,128 / 1,500
    66CLM v0.1 ยท 8B34.92%419 / 1,2001,200 / 1,200

    Switch between text, image, and overall accuracy.

    Full leaderboard and evaluation details โ†—Calibration and I/C indices โ†—

    JevBench and Decision Index

    Decision Index ยท JevBench Public 231 ยท JevBench Hard 111
    ModeleffortDecision Index โ†‘0.2.1JevBench Public 231JevBench Hard 111
    Accuracy โ†‘ECE โ†“Accuracy โ†‘ECE โ†“
    SPX-CD-Pro158.3089.61%207 / 2313.46%79.28%88 / 1114.73%
    2โ€”90.04%208 / 2315.78%80.18%89 / 1119.80%
    SPX-CD-Flash154.6588.31%204 / 2313.36%77.48%86 / 1115.18%
    2โ€”89.18%206 / 2312.41%77.48%86 / 1117.06%
    Jev 1.13.0โ€”57.9186.58%200 / 231โ€”72.97%81 / 111โ€”
    Decision Index โ†—JevBench โ†—

    05 / OVERVIEW

    Decisions your code can use.

    Provide the context, a question, and candidate options. SPX-CD returns a structured result that your application can use to choose an action, select multiple options, or apply a confidence threshold.

    In a browser agent, SPX-CD selects the next action. Browser tools click, type, and read the updated page before the model makes another decision.

    SPX-CD-Flash, SPX-CD-Pro, and SPX-CD-Omni are open source on Hugging Face, with LoRA weights, inference code, and evaluation results.

    06 / BUILD

    One interface. Four decisions.

    choice

    Choose one candidate action.

    multi_choice

    Select multiple options within the requested limits.

    noul

    Estimate the probability of yes.

    score

    Score an ordered set of levels.

    Try SPX-CD.

    Run a task in the Playground or connect your application through the API.

    Playground โ†—Read the API docs โ†—

    SURD AI / PLAYGROUND

    Create an account โ†— Sign in

    Continue in the platform Playground with your account allowance.