Jev directory
CommunityJev Models & Reimplementations
Tev1 on Ollama
Together AI's experimental decision family on Ollama, fine-tuned from Qwen3.5 at 4B and 0.8B: 93K downloads, up to 64 questions per call, an 88.0% development accuracy on 1,000 decisions, and MIT-licensed dataset builders and training scripts.
- Reading time
- 3 min
- Published by
- Community
- Author
- Together AI
- Last verified
- Oct 2026
- Platforms
- CLI
- Capabilities
- Open sourceSelf-hostedRuns locally
Highlights
- Experimental decision family from Together AI, fine-tuned from Qwen3.5 in 4B and 0.8B sizes, with 93K downloads on Ollama.
- The 4B model is the more accurate of the two; the 0.8B model is an 812 MB download for tight memory budgets.
- Together AI's development numbers for Tev1 4B at temperature 0 with thinking off: 880 of 1,000 on the main decision set and 300 of 300 on policy transfer.
- Training: 37,840 examples in one state, question, options format, from MultiNLI, BoolQ, Banking77, AG News, SST-5, policies, routing, and a taxonomy.
- Dataset builders and training scripts are MIT licensed; the page advises staying within 2 to 24 options, the range used in training.
Watch out
The 88.0% and 100% figures are Together AI's own development numbers, run on the same eval mix used while the model was built, so they are not independent; the policy transfer set is synthetic. The MIT licence covers the dataset builders and training scripts, no weights licence is stated, and prompt injection, non-English inputs, and calibration were not fully tested.
Questions about this entry
Short answers drawn from the linked project or page.
A family of experimental decision models from Together AI, fine-tuned from Qwen3.5 in 4B and 0.8B sizes. Given a question and options, it selects the answer based on likely outcome and returns a probability for each option.
Use ollama pull tev1:4b for the 4B model or ollama pull tev1:0.8b for the smaller one; Ollama 0.35 or later is required. The library lists 13 model entries, including the latest aliases at 256K context.
Together AI's development run on the 4B model at temperature 0 with thinking off scored 880 of 1,000 on the main decision set, 300 of 300 on policy transfer, and 1,300 of 1,300 valid single-letter answers. The page says the eval mix was used during development, so it is not independent.
37,840 examples converted into a shared state, question, options format: MultiNLI 5,000, BoolQ 3,000, Banking77 3,000, AG News 1,500, SST-5 2,000, programmatic policies 13,500, routing 6,000, and a research taxonomy 3,840. Tev1 takes inspiration from Jev but no training data came from it.
The page says the dataset builders and training scripts are MIT licensed; it does not spell out a licence for the model weights themselves. The 0.8B model is an 812 MB download and the 4B checkpoint is 4.4-4.6 GB.
Keep inputs short: every question is scored with the full state and question set in its prompt, Tev1 runs with a context of about 2,000 tokens, and the longest training example is about 1,500 tokens. Request bodies can be up to 64 KiB.
The page warns it can be wrong and should not be the only check on a high-stakes decision, that you should add a none option if nothing else might fit, and that Together AI has not fully tested prompt injection, languages other than English, calibration, or inputs unlike its training data.
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
Bespoke Labs' 9B decision model on Ollama, fine-tuned from Qwen3.5-9B: Apache-2.0, up to 64 typed questions per call, answers in under 100 ms on a MacBook Pro with an M5 Max, and 75.7% mean accuracy across 13 human-labelled datasets.
Jev Models & Reimplementations#community#open-models#local
Community
The Ollama build of iapp's OpenThai-SystemOne: a 0.8B Thai and English decision model whose v0.3 weights are fine-tuned for Ollama's /v1/systemone letter readout, reporting a 74.3 macro across 13 public subsets for Q8_0.
Jev Models & Reimplementations#community#open-models#local
Community
Cloudflare's 9B multimodal decision model on Ollama: text, JSON, or images in one non-autoregressive forward pass, Apache-2.0, and the lowest measured latency of any decision model in Cloudflare's Decision Index runs at 38.8 ms median.
Jev Models & Reimplementations#community#open-models#local
Community
ActBro's 2B calibrated System 1 decision model on Ollama, built on OpenBMB's MiniCPM5-2B: 78.79% on JevBench among sub-2B open-weight models, 86.41% on the 10-source decision-v7 set, and one output token per decision via /v1/systemone.
Jev Models & Reimplementations#community#open-models#local
Community
An unofficial cross-encoder fine-tune of ModernBERT-ja-310m for Jev Local: it reads a question-plus-state and one candidate per sequence and hits 92.62% on the JGLUE JNLI test, with CC BY-SA 4.0 weights.
Jev Models & Reimplementations#community#open-models#local
Community
The 322M mmBERT-base build of Laya for 100+ languages: state plus typed questions return typed answers with probabilities in one forward pass, with up to 8,192 tokens for long documents.
Jev Models & Reimplementations#community#open-models#local
Community
From 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.