[Critical] K2.6 model overloaded – unusable under normal load

What version of Kimi Code CLI is running?

latest

Which open platform/subscription were you using?

Kimi Code (Allegretto membership)

Which model were you using?

K2.6

What platform is your computer?

macOS (Apple Silicon)

What issue are you seeing?

K2.6 model constantly returns the error:

"The engine is currently overloaded, please try again later"

This happens frequently (multiple times per hour), even though:

• quota is available
• rate limits are not exceeded
• account is active (paid membership)

The issue occurs both in Console and CLI.

This makes the service unreliable for any continuous workflow.

What steps can reproduce the bug?

1. Send a request to K2.6 (Console or CLI)
2. Repeat several times within short time window
3. Error appears randomly

No specific prompt required.

What is the expected behavior?

Model should respond normally without overload errors under standard usage conditions.

Additional information

No response

The report identifies the Console and CLI as affected entry points but names no repository file or test. Start by tracing how K2.6 requests and overload errors are handled, then reproduce the failure under repeated standard requests; done means normal usage no longer produces unexplained overload errors, with regression coverage.