**Local text classification, workflow routing, and fixed-choice scoring with JevStyle.** This 0.8B v3 release is the package’s default text decision model. It accepts text and JSON, not images; its PyTorch, MLX, and GGUF distributions require the matching decision runtime. Maintained by **Chaoliang Yan** (GitHub: `lawrence3699`), as an independent project, not affiliated with TypeSafe AI. Training provenance is specific to each release; see this card’s training section. The separate 2B v4 release discloses soft-label distillation from TypeSafe Jev 1.14 through its public API. Choose a workflow: local text classification in Python · choosing a local decision model · this model’s setup and limits · evaluation evidence.

翻訳待ち・原文を表示

テキスト分析

入力テキスト→ 出力テキスト

料金未確認マルチモーダル・チャネル料金は個別に確認

詳細を見る ↗

比較に追加出典・文書 ↗

Jev-Style-2B-Decision-v3

chaoliangUNSW

基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。

テキスト分析

入力テキスト→ 出力テキスト

料金未確認マルチモーダル・チャネル料金は個別に確認

詳細を見る ↗

比較に追加出典・文書 ↗

Jev-Style-2B-Decision-v3-GGUF

chaoliangUNSW

> **Hosted API status · 2026-10-28:** The hosted API is temporarily unavailable for maintenance. For a working first try, open the browser demo (2B v4), or run the open weights locally. Service information. **Quickest start: the jev-style package** (0.3.0 or later) downloads these files and serves a local `/v1/systemone` API. It needs this repository's scorer, `jev-score-v2`, built once from `build_jev_score.sh` (see the Quick start below): **Jev-style decisions, now at 2B, for llama.cpp.** These are the GGUF builds of Jev-Style-2B-Decision-v3: F16, Q8_0 and a **1.27 GB** Q4_K_M, with a small bundled scorer (`jev-score-v2`) that reads the decision at one verdict slot per option. Full results, protocols, training data and licences are on the main model card.

翻訳待ち・原文を表示

テキスト分析

入力テキスト→ 出力テキスト

料金未確認マルチモーダル・チャネル料金は個別に確認

詳細を見る ↗

比較に追加出典・文書 ↗

Jev-Style-2B-Decision-v3-MLX

chaoliangUNSW

基础档案已收录，详细介绍后续补充。可先查看厂商、分类与来源。

基本情報を収録・詳細は順次追加

テキスト分析

入力テキスト→ 出力テキスト

料金未確認マルチモーダル・チャネル料金は個別に確認

詳細を見る ↗

比較に追加出典・文書 ↗

Jev-Style-Qwen3.5-2B-Decision-GGUF

chaoliangUNSW

> **Jev-Style v3 is available — smaller and stronger:** Jev-Style-0.8B-Decision-v3-GGUF scores **79.2%** on the 2,000 typed decisions (v1: 53.4%, v2: 73.5%, as reported on the v2 card), takes **25,600-token** inputs, works across **51 languages** and scores options without the 26-letter cap (tested with 77 options), all at **0.8B** parameters. This repository preserves v1; v2 is here. > **Hosted API status · 2026-10-28:** The hosted API is temporarily unavailable for maintenance. For a working first try, open the browser demo (2B v4), or run the open weights locally. Service information. A **Jev-style decision model**: it does not write text. Give it a state, a question and a list of options, and it returns the decision **with calibrated probabilities** from a single token position. Runs in **LM Studio** and **llama.cpp**.

翻訳待ち・原文を表示

テキスト生成

入力テキスト→ 出力テキスト

料金未確認マルチモーダル・チャネル料金は個別に確認

詳細を見る ↗

比較に追加出典・文書 ↗

Jev-Style-Qwen3.5-2B-Decision-MLX-bf16

chaoliangUNSW

> **Jev-Style v3 is available — smaller and stronger:** Jev-Style-0.8B-Decision-v3-MLX scores **79.2%** on the 2,000 typed decisions (v1: 53.4%, v2: 73.5%, as reported on the v2 card), takes **25,600-token** inputs, works across **51 languages** and scores options without the 26-letter cap (tested with 77 options), all at **0.8B** parameters. This repository preserves v1; v2 is here. A **Jev-style decision model**: it does not write text. Give it a state, a question and a list of options, and one forward pass returns the decision **with calibrated probabilities** - in 77 ms on an M1 Max. GGUF builds (BF16 / Q8_0 / Q4_K_M) for LM Studio and llama.cpp: chaoliangUNSW/Jev-Style-Qwen3.5-2B-Decision-GGUF

翻訳待ち・原文を表示

テキスト生成

入力テキスト→ 出力テキスト

料金未確認マルチモーダル・チャネル料金は個別に確認

詳細を見る ↗

比較に追加出典・文書 ↗