How to pick Kimi K3 with 1M context?

Support Help
,
1

I need a large context window for planning work. A 200k context leads to summarization too frequently. I typically use around 300–370k tokens, and that works well for me.

Is there any way to use Kimi K3 with the 1M context window in Cursor?

Also, I’m curious why this mode was removed. I’m not sure what the rationale was.

5

Hey, thanks for the request. I’ll break it down point by point.

First, about Max Mode. It wasn’t really “removed”, it was replaced. Before, it was the only way to say “I want a large context window”. Now context size is a separate setting in the model picker, so you can select it directly (for example, Sonnet, Opus, GPT-5.6, and Gemini offer options up to 1M). This does not affect pricing. Colin explained it here: What does Cursor removing “Max Mode” mean for usage-based pricing users?

Now, the honest part about Kimi K3. For models with a large physical window, the default working window in normal mode is 200K (that’s why you see summarization). You get the full window via the Context option in the picker. The issue is that for Kimi K3, this Context option is not currently shown in the picker, only the reasoning choice is. So on usage-based pricing there’s no direct way right now to select its full 1M context in the IDE. That’s not ideal, I’ve passed it to the team as a request to add context selection for Kimi K3 like the other 1M models.

What you can do right now:

  • If you need a big context specifically for planning, pick a model that has the 1M option in the picker (Sonnet, Opus, GPT-5.6, Gemini). You choose the context size directly in the model picker.
  • Alternatively, run the task via Cloud Agent. It uses the model’s full window (availability depends on your plan).

I can’t share an ETA for adding context selection for Kimi K3, but if there’s an update I’ll post it in the thread.