Early access A higher-volume API, and Decido fine-tuned on your own decisions. Request access →
yesp 0.97
Introducing

Decido

Fast, accurate decision models for AI agents. Runs on your laptop.

Ask a yes/no or multiple-choice question about any text. Get the answer and how sure it is. Open models, more accurate than Jev 1.13, the leading decision API.

Accuracy of public decision models, six-benchmark mean (%)
ahead of Jev →DECIDOOTHER PUBLIC MODELSDecido-1-Ultra: 62.16 (laptop CPU)62.2 Decido-1-UltraDecido-1-Fast: 73.01 (laptop CPU)73.0 Decido-1-FastDecido-1: 85.38 (laptop CPU)85.4 Decido-1Decido-1-Think: 88.71 (GPU)88.7 Decido-1-ThinkDecido-1-Max: 89.23 (GPU)89.2 Decido-1-MaxLaya: 58.2458.2 LayaKev-4B: 74.2574.3 Kev-4BOpen-Jev-9B: 75.6775.7 Open-Jev-9BNeoHorse-Jev-4B: 77.7077.7 NeoHorse-Jev-4BJEV-27B: 84.07 (distilled from Jev)84.1 JEV-27BJev 1.13: 83.8583.9 Jev 1.135560657075808590tiny & lightning fast← our best← the one to beat
← swipe for the smaller models
FIG. 1 Mean accuracy on the six benchmarks Jev 1.13 is evaluated on. Decido, our measurements (Max and Think on GPU; Decido-1, Fast and Ultra on a laptop CPU). Published or self-reported figures (model cards and the Jev Decision Index, October 2026). Details →
Try it live → pip install decido-ai
launched November 2026
Accuracy

Ahead of the leading decision API, without a GPU.

Decido-1 is a mixture-of-experts model with 4B active parameters, quantized to 8 bits. It scores higher than Jev 1.13 on four of the six benchmarks and on the mean. Decido-1-Max, which reasons before it answers, reaches 89.2 and is ahead on all six, in one set of open weights.

Per-benchmark results →
Decido-1 (laptop CPU)Jev 1.13 (published)707580859095JevBench: Decido-1 87.61 · Jev 1.13 87.18JevBenchKev: Decido-1 84.35 · Jev 1.13 85.52KevOpenJev: Decido-1 78.37 · Jev 1.13 72.96OpenJevNimble: Decido-1 87.23 · Jev 1.13 91.84NimbleVitaminC: Decido-1 86.14 · Jev 1.13 78.46VitaminCMASSIVE: Decido-1 88.57 · Jev 1.13 87.14MASSIVEMean: Decido-1 85.38 · Jev 1.13 83.85Mean← ahead on average
FIG. 2 Accuracy (%) per benchmark. Jev 1.13 figures are as published; we have not run Jev on our harness.
Speed

Five questions in 0.13 seconds, on a laptop.

Decido-1, Fast and Ultra read the answer straight from the model's probabilities: one pass over the text, nothing generated and nothing to parse. That makes them fast enough to check every step of an agent loop. Decido-1-Max and Decido-1-Think reason first, for the hardest calls.

0 bytes private by defaultsent to anyone's cloud when you run it locally. Decido runs on your laptop, server or air-gapped machine, with no per-call fees.
Reliability

The same answer, every time.

On CPU, Decido's outputs are bit-identical across calls and request history (at a fixed thread count), so any decision can be replayed exactly for an audit. It is also far less sensitive to the order in which answer options are listed.

Probabilities are calibrated, so a threshold can be set like an error budget: act automatically when the model is sure, and ask a person when it isn't.

Answers that change on a repeat callAnswers that change on a repeat call: Decido-1 0.0%0.0% Decido-1Answers that change on a repeat call: Jev 1.13 4.3%4.3% Jev 1.13Answers that flip when options are reorderedAnswers that flip when options are reordered: Decido-1 4.8%4.8% Decido-1Answers that flip when options are reordered: Jev 1.13 14.6%14.6% Jev 1.13← literally zero
FIG. 3 Lower is better. Decido-1 on CPU; Jev 1.13 published figures (measured with 64 options; ours on each benchmark's own options).
How it works

A document and typed questions in. Probabilities out.

1Describe the situationAny text or JSON: a support ticket, a tool call, a claim, an agent's next step.
2Ask typed questionsYes/no, pick one option, or a score on a scale, written in plain English. Many questions per call.
3Act on the probabilitiesEach answer comes with a probability for every option. Set a threshold to act, block or ask a person.

copy, paste, done ↓

# pip install decido-ai
from decido import Decido

ticket = "Hi, I was charged twice for my March invoice. Can you refund the duplicate?"
m = Decido("decido-ai/decido-1-fast-1.7b")        # downloads once, then runs locally
r = m.decide(ticket, {
    "refund":  "Is the customer asking for a refund?",
    "team":    {"q": "Which team?", "options": ["billing", "technical", "sales"]},
    "urgency": {"q": "How urgent?", "scale": ["low", "medium", "high"]},
})
# r["refund"] -> {"yes": 0.98, "no": 0.02}    r["team"] -> {"billing": 0.95, "technical": 0.02, "sales": 0.03}

if r["refund"]["yes"] >= 0.95:
    issue_refund()                                   # sure: act (your own functions)
else:
    send_to_human()                                  # unsure: ask a person
decido models                                         # the six models, scores and latency
decido run decido-ai/decido-1-fast-1.7b "Hi, I was charged twice for my March invoice." \
    -q refund="Is the customer asking for a refund?" \
    -q team="Which team?|billing,technical,sales" \
    -q urgency="How urgent?|scale:low,medium,high"
decido serve decido-ai/decido-1-fast-1.7b --port 8080    # local API, Jev 1.13 wire format; keeps running

# in another terminal:

curl localhost:8080/v1/systemone -H "content-type: application/json" -d '{
  "state": "Hi, I was charged twice for my March invoice. Can you refund the duplicate?",
  "questions": {
    "refund": {"type": "noul", "instructions": "Is the customer asking for a refund?"},
    "team": {"type": "choice", "instructions": "Which team?",
             "criteria": {"billing": "", "technical": "", "sales": ""}}
  }
}'
# -> {"answers": {"refund": {"type": "noul", "noul": 0.98},
#                 "team": {"type": "choice", "choice": "billing", "confidence": 0.92,
#                          "probabilities": {"billing": 0.95, "technical": 0.02, "sales": 0.03}}}, ...}
# Existing Jev SDK clients work unchanged: TypeSafeClient(base_url="http://127.0.0.1:8080", api_key="local")
# (the SDK zstd-compresses large requests; for decido serve to accept them: pip install "decido-ai[server]")
Open source (Apache-2.0; Decido-1-XL: MIT): decido-ai on PyPI · models on Hugging Face. One pip install on Mac (Apple Silicon), Linux (x86_64, aarch64) and Windows 10+ (x86_64): nothing to compile. Quickstart → On a GPU: pip install "decido-ai[gpu]" and device="cuda" (Decido-1-Max: CPU and Mac GPU for now; GPU weights soon).
Models

Six open models, one interface.

lightning fast!
Fastest

Decido-1-Ultra

Lightning fast and featherweight: 0.6B parameters, about 4 GB of RAM. Five questions in 0.13 s.
Benchmark mean
62.2
5 questions, 1k tokens
0.41 s CPU · 0.13 s Mac GPU
Hardware
laptop · ~4 GB RAM
Deterministic
bit-identical
Fast

Decido-1-Fast

Lightning fast and light: 1.7B parameters, about 8 GB of RAM. Built for checks inside agent loops.
Benchmark mean
73.0
5 questions, 1k tokens
0.64 s CPU · 0.30 s Mac GPU
Hardware
laptop · ~8 GB RAM
Deterministic
bit-identical
Accurate on CPU

Decido-1

Mixture of experts, 4B active parameters. More accurate than the leading decision API, on a laptop.
Benchmark mean
85.4
5 questions, 1k tokens
4.1 s CPU · 1.1 s Mac GPU
Hardware
laptop · ~30 GB RAM · GPU
Deterministic
bit-identical
Reasons first

Decido-1-Think

Decido-1 that reasons before answering. Ahead of Jev 1.13 on all six benchmarks.
Benchmark mean
88.7
Per question
~19 s CPU · ~10 s Mac GPU
Hardware
laptop · ~30 GB RAM · GPU
Deterministic
0% on CPU and Mac GPU
Our most accurate

Decido-1-Max

Reasons before answering, for the hardest calls. Ahead of Jev 1.13 on all six benchmarks.
Benchmark mean
89.2
Per question
~28 s CPU · ~11 s Mac GPU
Hardware
laptop · ~30 GB RAM (GPU path coming)
Calibration
Brier ≤ Jev on 5/5
Largest

Decido-1-XL

Our largest model, for GPU servers.
Benchmark mean
87.3
Hardware
4× A100 80 GB
Install
decido-ai[xl]

How each model was evaluated: see the technical report.

Research

How it performs, and where it falls short.

yes, we list the weak spots too

Early access

Own your agent's decisions.

We're working with a small number of design partners building AI agents. We fine-tune Decido on your own decisions and you run it on your own servers.