Starnexus gives you direct access to the leading models from Xiaomi, Moonshot, Qwen, Z.ai, and Google. Pick the right one for each task.
Xiaomi's MiMo models read text, images, audio and video, with a 1M-token context.
A 1M-context model at the lowest price we run.
Xiaomi's fast native multimodal model. It reads text, images, audio and video, with a 1M-token context for long documents and multi-step tasks.
Xiaomi's flagship, and the cheapest flagship we run.
Xiaomi's most capable model, built for complex software engineering and long, multi-step work. It reads text, images, audio and video, with a 1M-token context.
Xiaomi's flagship, served faster.
MiMo V2.6 Pro on a faster serving tier, for long multi-step work where waiting on each step adds up. Included on the Pro plan.
Built for length. Kimi reads and holds more at once than any other lab here, so it gets the jobs where the whole document matters: long reports, large codebases, hours of transcript.
Moonshot's flagship. Reads more at once than anything we carry.
Moonshot's most capable model, and the best in our lineup at holding a very long context: whole codebases, long reports, hours of transcript, without losing the thread. Included on the Pro plan.
One of the most widely used model families in the world, and the most even across everyday work: drafting, editing, summarising, and translating. Strong in many languages, and quick enough to use all day.
Fast, cheap, and reads images.
The quick end of the Qwen line, with vision and a long context window. Included on the Free plan.
The strongest Qwen, for writing and everyday work.
A sparse mixture-of-experts model that reads images as well as text, with a 1M-token context for long documents and multi-step work. Alibaba puts it at 2.4 trillion parameters. Included on the Pro plan.
The reasoning specialists, and the agents. GLM leads our lineup on multi-step agentic work and posts the highest competition-maths scores of any lab here, so it gets the tasks that take many steps and the questions that need working out.
Z.ai's flagship. Leads on agents and maths alike.
Reasons at the level of the largest models while answering far quicker than most of them, at a fraction of the cost. A strong default for hard questions, long documents, and code.
Z.ai's fast, low-cost model.
Reads images alongside text, with a context of about 1.3M tokens. It always reasons before it replies; Z.ai's API has no way to turn that off.
The multimodal line. Gemini reads images, audio, and video alongside text, with a very long context window, so it handles screenshots, recordings, and large documents in one go.
A fast, cheap default.
Google's fastest model, tuned for instant answers rather than deep thought. Handles everyday tasks and complex agentic workflows with multimodal input. A strong default for most chats.
Extended thinking over the same multimodal core.
Extended thinking mode over the same multimodal core as the rest of the Gemini family: text, images, video, audio, and code, with a 1M-token context. Optimized for software engineering, agentic workflows, and multi-step reasoning.