Your everyday decisions don't need a big LLM.
Is this ticket billing or bug? Would on-call act on this log line? Which file first? Teams are moving these small yes/no and pick-one calls onto a small model they own: it runs on a plain CPU (8-bit, 598 MB, about 1.6 GB RAM loaded), next to their production, and their data never leaves.
Try it on your own data or see how to get your own
Where it fits in your company
Pick-one routing is the shape of decision the open model handles well today. When it is not confident it says unsure instead of guessing; training on your own examples is how that gap shrinks on your decisions.
Try it now, with your own lines
Privacy: what you paste is used only to answer this request. It is not logged or stored; we keep only anonymous counts (visits, runs, clicks). Don't paste secrets or personal data anyway: this is a public demo on our machine. Demo only: the paid product never runs on our servers; your model runs on yours. Limits: 6 lines, 5 runs per 10 minutes.
Get a model that knows your rules
Everything is open source (Apache-2.0): the server, the training scripts, the base model. Prepare your data, train on any GPU, quantize to 8-bit ONNX, run it in Docker. github.com/muthuishere/openjevx
Yes → we train it on our GPUs from your documents and records, and hand you the model, a way to run it, and a report on your own held-out cases. Retrain monthly. Status: pilot. We run the pipeline by hand for the first few teams.
A packaged appliance that trains in your own cloud account, so nothing leaves it. Status: planned (DigitalOcean first, then AWS). Ask for early access.
You always run the model. We charge for the hours we save you (data preparation, GPU work, checking it), never for the code. We never host your inference.
The report lists, for each decision: right and confident, confidently wrong, and abstained. We ship only when your own test cases pass, and we tell you plainly if they don't.
Pilot on your data
Send three decisions your team makes every day and where the data lives. We'll tell you if it fits, before anything is trained.