I’m using a third - party API (mimo - v2.5 - pro), and the agent scheduling of Cursor is extremely slow, so I’m thinking of switching to CC. There are no problems when using CC.
Steps to Reproduce
use mimo and ask questions，the first response is too slow if compare cc
Cursor Agent mode does multi-step work on every turn. It pulls context, searches your codebase, plans, runs tool calls, and often makes multiple model calls on top of the actual model API call. A direct CLI client like Claude Code usually sends one request, so time-to-first-token is naturally lower.
A few things to try if you want lower latency:
Use Ask mode for Q and A style questions. It has much less overhead than Agent.
Use Inline Edit with Cmd+K for targeted edits.
Keep Agent context small. Attach fewer files and use a smaller workspace.
For BYOK models, latency also depends on how the third-party endpoint performs under heavier multi-tool-call workloads, which is more work than a single chat turn.
If you’re seeing something more specific than general orchestration overhead, like long stalls before any output or specific errors, share a Request ID and we can take a closer look.