Skip to content

bug: incorrect tok/s shown for remote providers #6289

@Minh141120

Description

@Minh141120

Describe the Bug

The tokens per second (tok/s) metric displayed for the remote providers appears to be incorrect.
Observed values show around ~3 tok/s, but the actual response feels much faster.

Steps to Reproduce

1.Set up an API key for the Gemini or Anthropic (remote) provider.
2. Start a new chat.
3. Use the model gemini-2.0-flash-001 or claude-sonnet-4-0
4. Observe the displayed tok/s during message streaming.

Screenshots / Logs

2025-08-25.18-01-37.mp4
Image

Operating System

  • MacOS
  • Windows
  • Linux

Metadata

Metadata

Labels

Type

Projects

Status

No status

Milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions