Models and levels
Nouter groups its models into two levels: Fast and Thinking. In the model selector you choose a level, and the model that runs it is shown underneath. This page says which model that is today, so you can check it before you rely on it. If your organization doesn’t have levels yet, the selector shows the model’s name and this page still describes what runs.
- Fast: quick answers and everyday work.
- Thinking: reasons longer, for complex work. It uses more per request.
What runs each level
Section titled “What runs each level”| Level | Model | Maker | License | Takes | Effort | Usage |
|---|---|---|---|---|---|---|
| Fast | DeepSeek V4.1 Flash | DeepSeek (China) | Not verified yet | Text and images | No effort levels | 1× |
| Thinking | GLM 5.3 Flash | Z.ai (China) | MIT | Text and images | Low to Max (High recommended) | 1.5× |
- Maker is who trained the model, and the country where it is based.
- License is the model’s own. For DeepSeek V4.1 Flash we haven’t verified it yet, and we won’t say more than that until we have.
- Takes is what you can send it: both levels read text and images.
- Effort is how much the model reasons before answering. A model without effort levels runs the same every time; with levels, the one marked recommended is used unless you pick another.
- Usage is how much a level uses compared with Fast (1×), for a typical request that reads a long conversation and writes a short answer. It’s an estimate: your own requests will use more or less.
Where the models run
Section titled “Where the models run”Both models are open. We run them in the United States, through our inference providers, and the maker doesn’t receive your data.
Changes we have scheduled
Section titled “Changes we have scheduled”We announce a change here ahead of time, before it reaches your chats: at least 7 days before a level changes maker, and 30 days before a level’s price goes up.
No changes are scheduled right now.
History
Section titled “History”- 2026-11-02. Thinking moved from GLM 5.3 to GLM 5.3 Flash (same maker, Z.ai).
See also Which models your organization uses and Writing a request.