A System One model, also written System 1 model, is an AI decision model that returns typed answers and calibrated probabilities instead of generated text. You give it content and typed questions; it returns the selected option, a probability for every option, and a confidence value. Jev, from TypeSafe AI, was the first one, announced on 15 September 2026.
732 examples18 use cases8 recipes12 models
Every System One call is a state plus questions. A question is one of three shapes, and each returns a number you can branch on.
Commercial models you call through a hosted API, then open-source models you run on your own hardware. Prices, limits and sources for each.
Jev 1.13 is the first System One model. It reads text you supply, answers typed questions about it with calibrated probabilities, and returns no generated prose. TypeSafe AI shipped it on 15 September 2026 as a closed managed API.
Decider 1 is meraGPT's System One model. It answers typed Choice, Score and Noul questions with probabilities over a 4,096-token request and writes no prose. It has been on meraGPT's hosted API since 22 September 2026, with no public weights.
Solar Decide is Upstage's System One model, served on Solar Mini 4 with a 512K-token context. It answers typed Choice, Score and Noul questions with probabilities, writes no prose, and has been in beta on Upstage's API since 22 September 2026.
Tev1-4B-experimental is Together AI's Jev-like classifier, a Qwen3.5-4B fine-tune that reads a state and a set of 2 to 24 options and returns one answer letter. Together released it alongside the training recipe and a tutorial on 23 September 2026.
Span-01 is Respan's behavior classifier for agent traces. It reads a conversation span and behaviors you define in plain language, and returns the probability that each is present, absent or not observable. It costs $0.02 per million input tokens, and a free Lite tier exists.
d1 is Liquid AI's first decision model. It answers typed Choice, Score and Noul questions with probabilities and writes no prose. It has been on Liquid's API as d1:free since 29 September 2026, with no paid pricing published yet.
OpenAI's Decisions API is a hosted System One model built on GPT-6 Luna. You send text or images and a question with fixed answers, and it picks one. It has been in limited preview since 29 September 2026, with no price or docs published yet.
Mercury Decide is Inception's decision model, served as a System One endpoint. It answers typed choice, score and yes/no questions with a probability and writes no prose. It has been free on OpenRouter since 30 September 2026.
Kev is an open-source family of System One models from Jared Palmer, at 0.8B, 4B, 9B and 27B on Qwen3.5 and Qwen3.8 bases. It answers typed Choice, Score and Noul questions with probabilities, serves TypeSafe's /v1/systemone API, and runs on your own GPU under Apache 2.0, with the 0.8B to 9B models also on a Mac. Kev-4B is also on OpenRouter at $0.042 per million input tokens.
Laya is an open-weights System One model from Convai Innovations. It answers typed Choice, Score and Noul questions about text you supply, returns probabilities instead of prose, and runs on your own GPU or CPU under Apache 2.0.
CLM (Contrastive Language Models) is an open-weights System One model that scores a state against a set of candidate actions with a contrastive objective instead of generating text. CLM-8B serves a TypeSafe-compatible API and Jacky Kwok's team released it, its data recipe and its scaling-law study on 23 September 2026.
GLiNER2.5-Decide is Fastino Labs' open-weights System One model. A 340M-parameter encoder answers typed classification questions, extracts spans and relations, and enforces cross-decision rules, and Fastino shipped it on 24 September 2026 under Apache 2.0.
More vendors will appear here as they ship System One models. Know one? Submit it.
Projects, tools, cookbooks and write-ups from people building on these models.
Browser Use's fast browser agent where Jev picks the operation and the DOM element in a single request; a small LLM is only invoked to write text when typing is required. Demoed on a Google Flights search.
19k starsFlight search 7.1 s
Attempt to run a Jev-style System One decision model locally on a single RTX 3090. Companion site at openjev.com. The repository was renamed to TheoLeeCJ/SemIf on GitHub, and the old openjev URL now redirects to it.
TheoLeeCJOpenopenjev on github.com4.1k starsJSON vs direct 5.21x slowerPeak throughput 20.03 dec/s
Makes one AI trade decision every Monad block on the Kuru MON-USDC market using Jev. The public deployment runs in dry-run mode with a mock model.
jarrodwattsOpenjev-trader on github.com2.2k starsLoop, mock model p50 100 msDecision window ~300 ms/block
Official agent skill for Claude Code, Codex and other agent environments; installs via claude plugin marketplace add typesafe-ai/skills or npx skills add typesafe-ai/skills. MIT licensed.
Open reimplementation that trains a small model to choose among a changing list of text options, emitting one probability per option in a single forward pass. Ships Doom, chess and Wikispeedia demos.
Independent community collection of use cases, patterns, prompts and starter code for Jev, explicitly not official. It also links sibling collections by the same author and two related projects that report their own cost figures.
MCP server that gives Claude Code, Codex, Amp, OpenCode and other agents ten Jev judgment tools, each returning typed probabilities they can branch on. The author calls it early software.
320 starsPer call 150-500 ms
Walks through the Choice, Score and Noul primitives live in the TypeSafe playground, with speed and cost numbers and their caveats.
The patterns people reach for, grouped by the job they do.
Decide where a piece of work goes next and whether your code is allowed to act on it without a person.
Judge whether retrieved text is relevant, ranked correctly, or actually supports the claim it is cited for.
Screen inputs and outputs for injected instructions, policy breaches, and answers that disagree with themselves.
Turn messy text into typed fields, scores, and labels that downstream systems and models can use directly.
Pick the next move fast enough for a loop that is already running, whether that loop is a game, a robot, or an agent.
Code you can paste, with the thresholds left in your code.
An LLM cites a document for a claim. This recipe checks the citation in two steps: a plain string match that catches quotes missing from the source, then one noul question that returns the probability the quoted section actually supports the claim. Your code turns that probability into a verdict.
A single Choice question sorts an inbound support ticket into one of four queues and returns a probability for every option. The model supplies the label and the certainty; your router applies the thresholds, holds the doubtful tickets for a human, and keeps every side effect in your code rather than in the prompt.
A support ticket needs a category, a severity, and two yes/no facts that only matter for one category each. Asking in sequence costs four round trips and four copies of the ticket. This recipe sends all four questions in one call, including the speculative ones, and lets the routing code ignore what it does not need.
One Choice names the action the user is asking for and one Noul says whether the message confirms it. Two numbers come back, and a table of per-action thresholds in your code decides the rest: a read runs at moderate confidence, a refund needs more, and anything below the floor goes to a person.
Start here if the category is new to you.
A System One (System 1) model is a decision model that returns typed, calibrated answers instead of text. How it works, the name, and which models exist.
Updated
Jev is TypeSafe AI's decision model. It answers typed questions about text with calibrated probabilities, and it is not an LLM, because it writes nothing.
Updated
System One models return typed decisions; LLMs return text. A side-by-side on output, latency, cost, hallucination and when each one wins.
Updated
Engineers called Jev a zero-shot classifier within hours of launch. What the comparison gets right, what it misses, and what is actually new.
Updated
The eight most recent entries in the directory.
Send the link. Projects, tools, cookbooks, videos, write-ups and threads all count. Every entry is checked and credited to its author.
