Today we are resetting the model lineup. The entire previous generation is retired: the Xenon family, the CORe 6 and CORe 7 lines, SaferTitan, Agintek, and every experimental variant in between. Their entries have been removed from the platform.
This was a massive architectural change that required a rewrite from the ground up, and it happened because of human creativity. Our team sat down, questioned every assumption about how OpenCORe should answer people, and rebuilt the stack ourselves rather than iterating on what was already there. CORe EX1 is the result of that work.
The new lineup
Two models:
- CORe EX1 is the default. It is efficient, quick to respond, and built for everyday chat, writing, and general assistance.
- CORe EX1-XL is the larger option. Pick it for complex reasoning, long documents, and harder coding work.
Both models are always ready, so there is no cold start when you open a chat.
Why we did this
Craft. Building the lineup ourselves let us shape exactly how it thinks and answers, instead of accepting whatever was handed to us. Every choice, from personality to pacing, is ours.
What changes for you
If you use the chat app, you will see a short notice about the move the first time you load the page. Dismiss it once and it is gone. The model picker now offers EX1 and EX1-XL only.
If you build against the API, update your model IDs. The valid values are core-ex1 and core-ex1-xl. Requests naming a retired model now return model_not_found. The reasoning_effort and reasoning_budget parameters belonged to the old lineup and are no longer supported; setting them returns invalid_reasoning_budget. Everything else about the API, including streaming, tool calling, and authentication, works the same as before.
What stays
Streaming responses, live thinking summaries, chat history, memories, projects, sharing, and every other platform feature carry over unchanged. The only thing that changed is what answers on the other side.