Jev directory
CommunityJev Models & Reimplementations
Blitz on Replicate
karthiksync's Replicate model for structured choices: supply a case record as state, then ask one question with two to twenty named choices, or batch up to eight questions against the same state and get decisions with probabilities.
- Reading time
- 2 min
- Published by
- Community
- Author
- karthiksync
- Last verified
- Oct 2026
Used inJev for Software & DevTools
- Platforms
- API
Highlights
- Structured-choice model on Replicate: state carries the evidence, and choices is a JSON object of two to twenty option names and descriptions.
- Single decisions return decision, probabilities, and a truncated flag; batch requests return one result per question in results.
- Accepts up to eight questions about the same state, each with its own choices object and an optional id, through the questions array.
- Runs on Nvidia T4 GPU hardware with 7 public runs at capture; Replicate says it does not yet have enough runs for performance information.
- The readme includes a hypothesis-test example and a retry-helpfulness example, plus a note that truncated means some state text was dropped.
Watch out
The page states no licence, so neither weights nor commercial terms are established, and with only 7 runs recorded the model is effectively untested at scale. Blitz evaluates the evidence you supply rather than verifying it, and the readme itself says to check decisions and probability thresholds against your workflow before consequential use.
Questions about this entry
Short answers drawn from the linked project or page.
A Replicate model by karthiksync that makes structured choices from a supplied case record. It returns a selected choice and probabilities for the options you provide, either for one question or several against the same state.
Provide state with the case text, a question, and choices as a JSON string of two to twenty names and descriptions; choices and questions are JSON strings in the Replicate input. The web form starts with a complete example you can run or edit.
Yes, up to eight. Paste a JSON array into questions, each item with a question and its own choices object and an optional id; when questions is filled, the single question and choices fields are ignored.
A single decision returns decision, probabilities, and truncated; batch requests return one result per question in results. If truncated is true, some state text was omitted to fit the input limit.
No. The readme says it evaluates the supplied evidence and does not independently verify it, and it advises checking decisions and probability thresholds against your workflow before consequential use.
Nvidia T4 GPU hardware, per the listing. Replicate says it does not yet have enough runs of this model to provide performance information, and the page showed 7 public runs at capture.
The page does not state one. No license link, weights repository, or terms appear on the listing, so licence and commercial-use questions are unanswered by this source.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: Standard One SH (StandardThinking)
LinkedIn: What multimodal decision models actually change, and what they do not.
LinkedInJevDirectory.org
Hacker News: Decision Index 0.2.1 scores open reproductions of Jev on a frozen suite.
jackbravo4 pointsDecision Index 0.2.1 – 70 open reproductions of TypeSafe Jev's Decision ModelRead the thread on Hacker NewsHacker News: Pointer to a DiffusionGemma plus vLLM tutorial for a Jev-shaped classifier.
1 pointHow to build a Jev-style classifier with DiffusionGemma and vLLMRead the thread on Hacker NewsMore Jev resources like this
run-saina's 0.8B decision model for workflow automation on Replicate: context, question, and answers become structured probabilities for yes/no, single-choice, multi-choice, and rating questions, with distribution or thresholded decision modes.
Jev Models & Reimplementations#community#open-models#local
Community
Nace.AI's 34.6B decision model on Fireworks AI: state plus typed yes/no, multiple-choice, or score questions return a probability for every option in one forward pass, deployable on demand with a 131K-token context window.
Jev Models & Reimplementations#community#open-models#local
Community
A Qwen2.5-1.5B-Instruct fine-tune that emits typed dependency verdicts, safe, caution, or block with a recommendation, for Dependabot and Renovate pipelines, reporting 95.7% F1 on the card's 4,800-update OSV-Bench.
Jev Models & Reimplementations#community#open-models#local
Community
A local yes/no decision model that speaks Jev's wire format on a laptop CPU, refusing Choice and Score until those question types ship.
Jev Models & Reimplementations#community#python#open-models
Community
feder-cr/jevC++
feder-cr/jevC++An Apache-2.0 decision engine with a TypeSafe-compatible endpoint, document answers with source passages, and a ViZDoom demo that picks actions from structured game state.
Jev Models & Reimplementations#community#python#local
Community
deepanwadhwa/OpenDecisionPython
deepanwadhwa/OpenDecisionPythonA Python server in front of llama-server that reads one-token label probabilities and exposes them as POST /v1/systemone.
Jev Models & Reimplementations#community#python#open-models
Community
khimaros/verdictPython
khimaros/verdictPythonFrom the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.