CommunityJev Models & Reimplementations
Julia 1 WebGPU
A browser WebGPU export of Julia 1: the same 551 MB decision checkpoint, a Rust WASM tokenizer, and a JavaScript API that answers typed questions client-side with no server or key.
- Published by
- Community
- Author
- Supersonic Labs
- Last verified
- Oct 2026
Used inJev for Software & DevTools
- Platforms
- Browser
- Capabilities
- Open sourceSelf-hostedRuns locally
Highlights
- 100 real validation requests in Brave matched the original Julia 1 on 100 of 100 predictions.
- Maximum absolute logit difference was 0.00225, with a 5.84-second load and warmup from the browser cache.
- 100 decisions ran in a median 7.55 seconds after warmup — about 75 ms each — on the WebGPU path.
- model.onnx and model.onnx.data must stay together; strict encoding rejects over-budget inputs by default.
Watch out
Accuracy was not rerun in the browser: parity is output agreement on 100 requests, and near-tied logits can still flip under floating-point rounding. WebGPU availability and device memory vary by browser and GPU.
Questions about this entry
Short answers drawn from the linked project or page.
No server. The ONNX graph, the 551 MB external weights, and a Rust WebAssembly tokenizer load into the browser; a WebGPU-capable browser is required and the device needs memory for the full model. The page renders before the weights download, so press Load model to start.
Same weights and decision graph, exported to ONNX. The repository verifies output agreement rather than new accuracy: 100 of 100 predictions matched on 100 validation requests, with a maximum logit difference of 0.00225.
Run npm install then npm run dev, open the URL Vite prints, and select playground.html. The page renders before the model downloads; press Load model to fetch the ONNX graph, the 551 MB external weights and the Rust WebAssembly tokenizer and warm a WebGPU session.
predict(rows) returns indexes and display probabilities, and logits(rows) returns raw scores. Passing a list of rows batches several decisions in one call. Strict encoding is enabled by default.
100 decisions ran in a median 7.55 seconds after warmup, about 75 ms each, on the WebGPU path. Loading and warming from the browser cache took 5.84 seconds.
Provide exactly two options in false then true order, because noul is the boolean primitive. The playground treats one option per line and shows the selected answer, probabilities and raw logits.
model.onnx and model.onnx.data must remain together, since the graph references the external weight file. The browser package uses ONNX Runtime WebGPU to compile the operations to WGSL.
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 NewsMore Jev resources like this
A 400M-parameter Jev-compatible decision model with a browser build: 0.796 on typed decisions against Laya's 0.766, 17 ms per question on an L40S, and an ONNX file that runs in a tab.
Jev Models & Reimplementations#community#open-models#local
Community
1.3kGitHub stars
A local daemon that pulls open decision models and serves them on TypeSafe's /v1/systemone, so a Jev client switches hosts with TYPESAFE_BASE_URL.
Jev Models & Reimplementations#community#rust#open-models
Community
ollaya-dev/ollayaRust
ollaya-dev/ollayaRustThe Hugging Face hub for Laya's open System 1 checkpoints: the 421M English model at the repo root, a 322M multilingual build in a subfolder, and the typed-decisions fine-tune, all Apache-2.0.
Jev Models & Reimplementations#community#open-models#local
Community
Supersonic Labs' 144.3M-parameter decision model on mmBERT-small: choice, score, and noul answers from supplied options in one pass, 73.15% on 2,000 typed decisions, Apache-2.0.
Jev Models & Reimplementations#community#open-models#local
Community
An open distillation of Jev 1.13 onto Qwen3.8-27B: a 108.9M-parameter decision block answers typed questions in one prefill pass and adds a POST /v1/decide route beside a full System 2.
Jev Models & Reimplementations#community#open-models#local
Community
The 9B member of AutoTrust's Jev students: a 40.2M-parameter decision block on Qwen3.5-9B with 0.019 mean KL to Jev 1.13 and 70.7% HumanEval left untouched, small enough for one modest GPU.
Jev Models & Reimplementations#community#open-models#local
Community
From the Jev guides
Original write-ups on what this entry is for.
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
Hundreds of repositories carry the system-one tag, but the ecosystem is a dozen ideas in different sizes. A tour of the families, the licenses, and what a download does not include.
Jev Models & Reimplementations#open-models#hugging-face#local
Read article
From the Jev community
Posts from builders shipping with Jev right now.