### Check for existing issues

- [x] I have searched the existing issues and checked that my issue is not a duplicate.

### What happened?

When using claude code with thinking models (kimi-k2.5 for instance), it fails.

The same works ok with 1.81.12 version.

### Steps to Reproduce

1. Upgrade to the latest stable 2. Use claude code with thinker model 3. Requests fail.

### Relevant log output

```shell litellm.BadRequestError: Hosted_vllmException - {"object":"error","message":"19 validation errors:\n {'type': 'string_type', 'loc': ('body', 'messages', 2, 'ChatCompletionMessageGenericParam', 'content', 'str'), 'msg': 'Input should be a valid string', 'input': [{'type': 'thinking', 'thinking': \"The user wants to know what's in the mcpserver.py file. Let me read it for them. \"}]}\n {'type': 'literal_error', 'loc': ('body', 'messages', 2, 'ChatCompletionMessageGenericParam', 'content', 'list[union[ChatCompletionMessageContentTextPart,ChatCompletionMessageContentImagePart,ChatCompletionMessageContentVideoPart,ChatCompletionMessageContentAudioPart]]', 0, 'ChatCompletionMessageContentTextPart', 'type'), 'msg': \"Input should be 'text'\", 'input': 'thinking', 'ctx': {'expected': \"'…