CommunityJev Models & Reimplementations
AutoTrust JEV-35B
AutoTrust's bf16 System 1 decision model on Qwen3.5-35B-A3B with about 3B active parameters: Decision Index 0.3 public 59.73, ERQA 51.3% in one pass with zero robot training, 66.5 GiB of weights.
- Published by
- Community
- Author
- AutoTrust AI Lab
- Last verified
- Oct 2026
- Capabilities
- Open sourceSelf-hostedRuns locally
Highlights
- Built on Qwen3.5-35B-A3B with 35B parameters and about 3B active per token, reading up to 256 options in one pass.
- Decision Index 0.3 public suite 59.73 against JEV-27B-VL's 53.64, both scored with AutoTrust's own kit.
- ERQA 51.3% in one pass with zero robot training, level with PhysBrain 1.5's 52.8 and above every other open embodied model of 7-9B.
- 72 GB download and 66.5 GiB of bf16 weights; median single-request latency 204 ms on one B200.
- The card stresses JEV-35B is AutoTrust's independent open-weights model, not TypeSafe AI's closed Jev 1.13.
Watch out
The numbers are AutoTrust's own runs with the Decision Index kit, not a board entry, and the board's Full score also counts private tests only its maintainers run. ERQA's 400 questions carry roughly a ±5-point interval, and the repo ships a LICENSE file whose terms the card does not name.
Questions about this entry
Short answers drawn from the linked project or page.
Qwen3.5-35B-A3B, a mixture-of-experts model with 35B parameters and about 3B active per token. AutoTrust adds a System 1 decision path that reads up to 256 options in one pass, served through serve.sh with a POST /v1/decide API.
59.73 on the public suite in AutoTrust's own scoring with the kit, against 53.64 for JEV-27B-VL on the board. It answered all 140,178 scoreable requests with zero errors, System 1 only, every choice read in one pass.
No. The card says its decision training was text only, with no robot or embodied data, yet it scores 51.3% on ERQA in one pass, level with PhysBrain 1.5's 52.8 and above every other open embodied model of 7-9B (42.3-47.5).
66.5 GiB of GPU memory for the weights and a 72 GB download, so it targets a large GPU. The sibling NVFP4 build brings the same model down to 25.6 GB of download and 23.3 GiB of GPU memory for 32-48 GB cards.
No. TypeSafe Jev 1.13 is the hosted, closed model made by TypeSafe AI; JEV-35B is an independent open-weights model built by AutoTrust AI, and the card states it is not affiliated with, endorsed by, or a product of TypeSafe AI.
Median single-request latency is 204 ms on one B200 in bf16. On ERQA each question is one choice decision through POST /v1/decide, with a median 1.26 seconds per question at 32 parallel requests.
ERQA has 400 multiple-choice questions with about 26% chance accuracy, and the card notes the 95% interval of a score is about ±5 points. JEV-35B scores 51.2-51.3% in one pass; with thinking on, JEV-27B-VL reaches 61.5%.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
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 NewsHacker News: A one-token logprob wrapper that also classifies webcam frames.
allanrbo151 pointsA single function Jev-like wrapper for LLMs, including vision modelsRead the thread on Hacker NewsRelated terms
Glossary definitions related to this entry.
More Jev resources like this
The NVFP4 build of AutoTrust's JEV-35B: routed experts quantized to 4-bit cut the download to 25.6 GB and GPU weights to 23.3 GiB, keeping 96.6% of bf16's Decision Index answers.
Jev Models & Reimplementations#community#open-models#local
Community
AutoTrust's combined System 1 and System 2 bundle on Gemma-4-26B-A4B-it: Decision Index 0.2.1 of 58.05, a 24-slot head with noul, choice and score, and the base chat model left bit-identical.
Jev Models & Reimplementations#community#open-models#local
Community
alibiserikbay's Apache-2.0 Jev alternative on Qwen3.5-4B with a merged rank-16 LoRA: one forward pass returns a probability per option, 0.784 on JevBench's public hard tier at about 13 ms per decision.
Jev Models & Reimplementations#community#open-models#local
Community
The 9B JevK5: the same v0.3 data on Qwen3.5-9B, retrained for 1.5 epochs by a pre-registered sweep, reaching index proxy 0.794 and a 0.859 hand-written hard set, Apache-2.0.
Jev Models & Reimplementations#community#open-models#local
Community
chaoliangUNSW's v1 Jev-Style decision model in GGUF from 1.3 GB: calibrated choice, bool and score answers on Qwen3.5-2B-Base, 82.3% on five held-out task families with ECE 0.017.
Jev Models & Reimplementations#community#open-models#local
Community
An Apple Silicon MLX bf16 build of the v1 Jev-Style decision model: 77 ms decisions on an M1 Max with calibrated choice, bool and score probabilities, Apache-2.0 weights.
Jev Models & Reimplementations#community#open-models#local
Community
From the Jev guides
Original write-ups on what this entry is for.
Training recipes, looped architectures, and language ports: what the open side of the typed-decision class shipped this month, and how to pick by runtime and calibration instead of the leaderboard.
Jev Models & Reimplementations#open-models#training#local
Read article
Taking a decision checkpoint out of Python: what the GGUF and browser WebGPU exports preserve, what they change, and the parity evidence the model cards publish.
Jev Tools & CLIs#open-models#local#onnx
Read article
Interface, size, license, claims — in that order. A selection guide across the system-one catalog, from a 706K form scorer to 27B Jev students.
Jev Practices & Patterns#open-models#evaluation#benchmarks
Read article
From the Jev community
Posts from builders shipping with Jev right now.