Xenon 8 is our eighth generation of models and our biggest leap yet. It replaces the CORe 7 Series at the top of the lineup with three new models that think longer, reason more carefully, and understand more of what you give them. Every tier got a serious upgrade: deeper reasoning, sharper tool use, and a substantially higher parameter count across the board. We are keeping the exact counts to ourselves, but suffice it to say each Xenon 8 model is meaningfully larger than the model it succeeds.
Xenon 8 (X8): The Best We Have Ever Built
Xenon 8, or X8 for short, is the flagship and the most capable model we have ever shipped. It sits above everything else in the lineup. Compared to our previous flagship, it spends more effort working through hard problems before it answers, which shows up as better plans, cleaner analysis, and far fewer confident mistakes on long, multi-step tasks.
The bigger parameter budget gives X8 noticeably richer world knowledge and more reliable judgment in ambiguous situations. It is better at holding a huge codebase or a long conversation in its head at once, better at catching its own errors mid-thought, and better at saying "I'm not sure" when that is the honest answer. Multimodal understanding improves too: images, video, and documents are parsed with more detail and less hand-waving.
Use X8 when the problem is hard and the answer matters.
Xenon 8 Code (X8 Code): The Engineer, Upgraded
Xenon 8 Code, or X8 Code, takes everything we learned from CORe 7 Code and Agintek and pushes it further. It reasons about your repo as a system instead of a pile of files: it plans an approach, writes the change, runs the loop, and checks its own work before handing it back.
With a higher parameter count than its predecessor, X8 Code is sharper at architecture decisions, more careful with edge cases, and much better at long agentic runs where a small mistake compounds. Refactors stay consistent across more files. Debugging sessions waste fewer steps. It still knows when to ask and when to act.
Use X8 Code when you are shipping something real.
Xenon 8s (X8s): Speed, Smarter
Xenon 8s proves the small model does not have to be the dumb model. It is a compact, low-latency build of the Xenon 8 family with a higher parameter count than any fast model we have shipped before, tuned for rapid tool use and high-frequency loops where every millisecond matters.
X8s inherits the improved thinking of its bigger siblings in a distilled form: quick plans, fast answers, and the same refusal calibration and personality you already know. For CLI agents, interactive companions, and automation at scale, it is the new default for speed.
Use X8s when latency is the feature.
Better thinking, everywhere
The headline upgrade in Xenon 8 is how the models think. All three tiers were trained to reason more deliberately: breaking problems down, checking intermediate steps, and revising when something does not add up. In practice that means fewer hallucinated details, more honest confidence, and answers that hold up when you push back.
Bigger models, same promises
Every Xenon 8 model ships with a higher parameter count than the generation before it. We are not publishing exact figures; what matters is what you feel: better answers, better reasoning, better code. Everything else stays the same: the OpenAI-compatible API, multimodal input, tool calling, streaming, and the alignment stack the platform is built on. Existing integrations keep working. Just switch the model.
Availability
Xenon 8, Xenon 8s, and Xenon 8 Code are rolling out in the chat app and on the API starting today. The CORe 7 Series remains available for compatibility. If you are building against the API, the new models are available under the xenon-8 family of model IDs.