Read tool for kimi k2.5 on zen dose not work on images #8187

Closed
opened 2026-02-16 18:09:23 -05:00 by yindo · 1 comment
Owner

Originally created by @SirStisx on GitHub (Jan 31, 2026).

Originally assigned to: @fwang on GitHub.

Description

when kimi k2.5 use read tool on a image it says somthing like this,

"I can't tell what the image shows from just the base64 data - I don't have vision capabilities to decode and view it"

pasting a image in the message works,

swaping moddel in same conversation and telling to do the same works.

Plugins

No response

OpenCode version

v1.1.48

Steps to reproduce

  1. have a jpg in a folder
  2. tell kimi k2.5 to use read tool on the image

Screenshot and/or share link

No response

Operating System

mac os

Terminal

ghosty

Originally created by @SirStisx on GitHub (Jan 31, 2026). Originally assigned to: @fwang on GitHub. ### Description when kimi k2.5 use read tool on a image it says somthing like this, "I can't tell what the image shows from just the base64 data - I don't have vision capabilities to decode and view it" pasting a image in the message works, swaping moddel in same conversation and telling to do the same works. ### Plugins _No response_ ### OpenCode version v1.1.48 ### Steps to reproduce 1. have a jpg in a folder 2. tell kimi k2.5 to use read tool on the image ### Screenshot and/or share link _No response_ ### Operating System mac os ### Terminal ghosty
yindo added the bugzen labels 2026-02-16 18:09:23 -05:00
yindo closed this issue 2026-02-16 18:09:23 -05:00
Author
Owner

@github-actions[bot] commented on GitHub (Jan 31, 2026):

This issue might be a duplicate of existing issues. Please check:

  • #11304: Images returned from Read tool don't work with OpenAI-compatible providers (Kimi K2.5 specifically mentioned)
  • #11306: Read tool fails to read image files even with vision-capable models (paste works)
  • #11437: Read tool accesses JPEG file but vision model receives base64 string instead

These issues describe the exact same symptom - the Read tool returns base64-encoded images instead of letting vision-capable models handle them natively. This is particularly affecting OpenAI-compatible providers like Kimi K2.5 when accessed through Zen.

Feel free to ignore if your case is different, but these related issues may provide additional context or workarounds.

@github-actions[bot] commented on GitHub (Jan 31, 2026): This issue might be a duplicate of existing issues. Please check: - #11304: Images returned from Read tool don't work with OpenAI-compatible providers (Kimi K2.5 specifically mentioned) - #11306: Read tool fails to read image files even with vision-capable models (paste works) - #11437: Read tool accesses JPEG file but vision model receives base64 string instead These issues describe the exact same symptom - the Read tool returns base64-encoded images instead of letting vision-capable models handle them natively. This is particularly affecting OpenAI-compatible providers like Kimi K2.5 when accessed through Zen. Feel free to ignore if your case is different, but these related issues may provide additional context or workarounds.
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: anomalyco/opencode#8187