聊天回复前的决策助手：读取聊天内容判断对方意图和风险，一键生成草稿供你确认后发送，支持 Android、Windows、macOS、iOS。
单次前向传播即可对任意文本做类型化选择、打分和是非判断的非自回归决策引擎，支持100多种语言，并按请求自动路由到合适的检查点。
自托管的 AI Agent 工作台，把任务看板、Agent 编排和多端访问集中到一处，桌面、手机、浏览器或 API 都能用。
一站式大模型训练框架，支持 600+ 纯文本大模型和 300+ 多模态大模型的 CPT、SFT、DPO、GRPO 等训练，可用 LoRA 等 PEFT 方式或全参数微调。
一个能自主完成深度调研的智能体，可接入任意 LLM 提供商，自动搜索、抓取网页并汇总成研究报告。
Yes/no decisions on a laptop CPU in 28–130 ms: jev , one binary for Windows, Linux and macOS, now with jevos-v5 at 8-bit weights. 78.9% on the benchmark's headline yes/no
The first preview of the global edition of Jev: an English reply copilot for WhatsApp on Android, built by @smgonthebeat. It reads the chat on screen, works out what the
Yes/no decisions on a laptop CPU in 25–110 ms: jev , one binary for Windows, Linux and macOS, now with jevos-v4 at 8-bit weights. 87.0% on 821 new yes/no questions writte
Yes/no decisions on a laptop CPU in 25–110 ms: jev , one binary for Windows, Linux and macOS, now with jevos-v3 at 8-bit weights. 80.8% on 999 hand-written yes/no questio
…do in https://github.com/LLPhant/LLPhant/pull/513 Jev prompt injection detection by @f-lombardo in https://github.com/LLPhant/LLPhant/pull/514 Full Changelog : https://gi
Yes/no decisions on a laptop CPU in 25–110 ms: jev , one binary for Windows, Linux and macOS, with jevos-v2 at 8-bit weights. 80.3% on 999 hand-written yes/no questions,