Loading every model…
THE COMPLETE LIST
A System One model is a decision model: it reads a state and answers typed questions — choose an option, score on a scale, or give the probability that a condition holds — in one forward pass, with calibrated confidence and no generated text.
This page lists every System One model on the registry — 30 from 25 makers — with what each decides, its size, context, licence and price, and the benchmark numbers its publisher reports. Updated November 20, 2026.
| Model | Decides | Parameters | Context | Availability and price | Licence | Reported accuracy | Median latency | Released |
|---|---|---|---|---|---|---|---|---|
| vegaNandakishor M | choice, score, noul, classify, route | 800M | 72K tokens | Open weightsFree (open weights) | apache-2.0 | 55.6%His 11 October 2026 benchmark, published with its results and its harness (a notebook in vegaml): macro accuracy over 43 public tasks for the 0.8B, 0.717 for the 4B; with JevBench (231 items) and DecisionBench (1,075 rows), 3,841 decisions in all, seed 20261011. On his own set of 2,050 choice decisions over six intent workflows (esci, clinc150, appliances, banking77, when2call, apibank), one choice question each, which he has not published with its question wording, he reports 0.763 (calibration error 0.026) for the 0.8B with its public_intents adapter, 0.389 without the adapter and 0.803 for the 4B with it. He publishes no accuracy for the typed_decisions adapters | — | Oct 9, 2026 |
| ajevAndy Zhang | choice, score, noul | 26B | — | Open weightsFree (open weights) | apache-2.0 | — | 49 ms | Oct 5, 2026 |
| nativ-he-decisionYoav Pinto | choice, classify, route | 364M | 1K tokens | Open weightsFree (open weights) | cc-by-4.0 | 96.2%nativ-bench user-intent task, 4 options (MASSIVE Hebrew, 2,973 items; the maker's own suite) | — | Oct 4, 2026 |
| canopy-jevRuoxi Qiu | choice, noul, score | 27B | — | Open weightsFree (open weights) | apache-2.0 | 87.4%JevBench public set (231 items; 202 correct), the maker's own fresh zero-shot run with the upstream scorer | — | Oct 2, 2026 |
| docto-decisionBofeng Huang | choice, score, noul | 4.7B | — | Open weightsFree (open weights) | apache-2.0 | 88.9%Docto Decision Bench fr v0.1 (12 tasks, mostly silver labels; the maker's own benchmark) | 47 ms | Oct 2, 2026 |
| gutsyOmar (kouhxp) | choice, score, noul, classify, route | 800M | 8K tokens | Open weightsFree (open weights) | apache-2.0 | 73.2%JevBench public set (231 items; 169 correct), the maker's own run with Q8_0 | — | Oct 2, 2026 |
| waldORG2 AI | choice, score, noul, classify, route | 4.2B | 128K tokens | Open weightsFree (open weights) | apache-2.0 | 88.3%JevBench public set (231 items; 204 correct), maker's run with the JevBench CLI, v1.2, effort none; items used as a development scoreboard | 33 ms | Oct 1, 2026 |
| modernjev-decideMaziyar Panahi | choice, rank, route | 150M | 4K tokens | Open weightsFree (open weights) | apache-2.0 | 73.4%AgentToolDecisions-180K held-out test, next action type (1,158 decisions, 3 choices; the maker's split) | — | Sep 30, 2026 |
| startlux-decisionStartLux | choice, score, noul | 4.7B | 256K tokens | Open weightsFree (open weights) | cc-by-nc-4.0 | 88.3%JevBench public set (231 items; 204 correct), maker's run | 26 ms | Sep 29, 2026 |
| bekko-system-oneYuichi Tateno (hotchpotch) | choice, noul, score | 395M | — | Open weightsFree (open weights) | other | — | 5 ms | Sep 29, 2026 |
| biodecisionDimitris Papadopoulos | choice, score, noul, classify | 4.0B | — | Open weightsFree (open weights) | other | 75.3%BioDecision v2 held-out test set (63,432 overlap-screened decisions from public biomedical benchmarks; the maker's own split) | 39 ms | Sep 28, 2026 |
| bev-deciderAvishek Biswas | choice, score, noul | 478M | 2K tokens | Open weightsFree (open weights) | cc-by-nc-4.0 | 74.7%avbiswas/bev-decision test split (5,000 held-out questions over 2,617 states), the maker's own | — | Sep 28, 2026 |
| rizzo-flowRizzo AI Academy | choice, score, noul, classify, route | 4.0B | — | Open weightsFree (open weights) | apache-2.0 | 64.8%LocalLLaMA/typed-decisions test split (400 cases, 2,000 decisions), Q8_0 | 195 ms | Sep 25, 2026 |
| gliner2-5-decideFastino Labs | choice, score, noul, classify, extract, route | 340M | — | Open weights + hosted APIHosted, price not published, or free to self-host | apache-2.0 | 60.2%Fastino fast-decisions suite (17 datasets) | 38.3 ms | Sep 24, 2026 |
| deciderMapika | choice, score, noul, classify, route | 1.9B | 32K tokens | Open weightsFree (open weights) | apache-2.0 | 75.2%Decider regression set, 28 held-out tasks (decider-2b v11) | — | Sep 24, 2026 |
| julia-1Supersonic Labs | choice, score, noul, classify, route | 144M | 8K tokens | Open weightsFree (open weights) | apache-2.0 | 73.2%typed-decisions test set (400 cases, 2,000 questions) | — | Sep 23, 2026 |
| clmContrastive-LM | choice, score, noul, rank, classify, route | 8.0B | — | Open weightsFree (open weights) | apache-2.0 | — | — | Sep 23, 2026 |
| neohorse-jevTokenRhythm | choice, score, noul, classify, route | 4.0B | — | Open weightsFree (open weights) | apache-2.0 | 75.3%JevBench public set (231 items), vLLM, maker's run | — | Sep 23, 2026 |
| lumma-fevFrontiersMind | choice, score, noul | 649M | 8K tokens | Open weightsFree (open weights) | apache-2.0 | 64.0%typed-decisions (maker's table; split not stated) | 45.8 ms | Sep 22, 2026 |
| anyjevNokia | choice, score, noul, classify, route | 8.0B | — | Open weightsFree (open weights) | apache-2.0 | 77.1%LocalLLaMA/typed-decisions (2,000 held-out decisions), L2 with 300 labels per question | — | Sep 21, 2026 |
| kevJared Palmer | choice, score, noul, classify, route | 4.0B | 8K tokens | Open weightsFree (open weights) | apache-2.0 | 83.8%transfer-v4 (locked, out of domain) | — | Sep 19, 2026 |
| layaConvai Innovations | choice, score, noul, classify, route | 421M | 512 tokens | Open weightsFree (open weights) | apache-2.0 | — | 39.5 ms | Sep 18, 2026 |
| open-jev-deberta-v3-largeKotoba Labs | choice, score, noul, classify | 434M | 512 tokens | Open weightsFree (open weights) | apache-2.0 | 85.4%Kotoba held-out test, seen question types (banking77, SST-5, BoolQ) | 28 ms | Sep 18, 2026 |
| miraSAGEA | choice, score, noul, classify, route | — | — | Open weightsFree (open weights) | apache-2.0 | 61.3%s1-decision-bench | 30 ms | — |
| caliber-flex-5bonepass | choice, noul, score | — | — | Open weightsFree (open weights) | other | — | — | — |
| caliber-flex-1bonepass | choice, noul, score | — | — | Open weightsFree (open weights) | other | — | — | — |
| laya-typed-decisions-noulxpConvai Innovations | choice, score, noul, classify, route | 421M | 1K tokens | Open weightsFree (open weights) | apache-2.0 | — | 126 ms | — |
| laya-noulxp-cpuConvai Innovations | choice, score, noul, classify, route | 421M | 512 tokens | Open weightsFree (open weights) | apache-2.0 | — | 116 ms | — |
| laya-noulxp-gpuConvai Innovations | choice, score, noul, classify, route | 421M | 512 tokens | Open weightsFree (open weights) | apache-2.0 | — | — | — |
| laya-multilingual-noulxpConvai Innovations | choice, score, noul, classify, route | 322M | 1K tokens | Open weightsFree (open weights) | apache-2.0 | — | 44.9 ms | — |
Accuracy and latency are what each publisher reports, on its own suite and hardware. They are not a ranking: the suites differ. Parameters, context, licence and price are from each model’s manifest.
Pull any open-weight model from your terminal. Files are verified against their SHA-256 and cached.
pip install systemonemodels systemone pull convai-innovations/laya
Hosted models are called through their provider’s API; each model page has a playground that builds the request.
System One Studio fine-tunes Laya on your own labelled data, on your own machine, measures it against the base model on the same held-out split, and publishes it here in one step.
A System One model is a decision model: it reads a state and answers typed questions — choose an option, score on a scale, or give the probability that a condition holds — in one forward pass, with calibrated confidence and no generated text.
Here, on System One Models (systemonemodels.ai): a registry built only for System One models, like Hugging Face but for decision models. This page lists every one it knows about, open and hosted, and each model has its own page with files, a playground and its lineage.
The registry lists 30 System One models from 25 makers: 30 with open weights and 1 served as a hosted API, counting models that are both in each. People have published 22 fine-tunes and experiments on top of them.
laya, from Convai Innovations, released on Sep 18, 2026.
Open weights under a licence that allows commercial use: julia-1 (apache-2.0), laya (apache-2.0), vega (apache-2.0), mira (apache-2.0), laya-typed-decisions-noulxp (apache-2.0), laya-noulxp-cpu (apache-2.0), gliner2-5-decide (apache-2.0), laya-noulxp-gpu (apache-2.0), decider (apache-2.0), laya-multilingual-noulxp (apache-2.0), clm (apache-2.0), neohorse-jev (apache-2.0), rizzo-flow (apache-2.0), open-jev-deberta-v3-large (apache-2.0), wald (apache-2.0), lumma-fev (apache-2.0), kev (apache-2.0), anyjev (apache-2.0), canopy-jev (apache-2.0), nativ-he-decision (cc-by-4.0), docto-decision (apache-2.0), gutsy (apache-2.0), modernjev-decide (apache-2.0) and ajev (apache-2.0). Some other models publish weights under a non-commercial licence or none at all, so check each model's licence before shipping it.
By the median latency their publishers report, the fastest are bekko-system-one at 5 ms, startlux-decision at 26 ms and open-jev-deberta-v3-large at 28 ms. The figures come from different hardware and different request sizes, so measure on your own setup before choosing on speed.
Open-weight models are free to download and run on your own hardware.
There is no fair ranking yet. Every publisher reports accuracy on a different suite — their own held-out set, a JevBench tier, a domain benchmark — so the numbers in the table are not comparable with each other. The table shows the suite next to each figure; the reliable test is your own labelled examples.
Install the CLI with pip install systemonemodels, then systemone pull namespace/name downloads a model’s files, each verified against its SHA-256. Hosted models are called through their provider’s API, which each model page links to.
Yes. System One Studio fine-tunes Laya on your own data, on a Mac, Windows or Linux machine, and publishes the result here in one step. Any open-weight model can be fine-tuned with its own training code and published with systemone push.
Read on: AI decision models · Open-source Jev alternatives · the glossary · What is a System One model? · System One models vs LLMs · The landscape, in depth