OpenJev is not a Large Language Model (LLM). It's a different kind of AI. You can input large text to it, but it can only respond with multiple choice answers. YOU provide the multiple choice options plus the text you want it to judge.
It responds with all of your multiple choice options with a score of each for how well each option fits your prompt.
And it's FAST! CRAZY fast! My favorite LLM that I use for most things locally is qwen3-coder:latest (it's 17GB in size). It takes several seconds to respond for most prompts + the token generation time, which could be close to a minute in some cases.
OpenJev processed the EXACT SAME prompts at 5.75 PER SECOND! I'm not saying 5.75 tokens per second. I'm saying I feed it 5.75 LARGE ARTICLES per second and get 5.75 fully processed answers PER SECOND!
OpenJev is also about 17GB in size. The hardware I run it on is an RTX 3090 with 24GB of VRAM.
Note that it is NOT a FULL replacement for your LLMs. If you need CONTENT generated, you'll still have to use your LLMs. But if you can remake your prompts expecting just multiple choice responses, for THOSE you can use OpenJev.