Hi all,
I’ve been seeing some strange responses since yesterday when using the llama4 maverick model.
I have a tool attached named - `proceed_to_next_step` but instead of the model actually making a tool call, it gives the below text in response content.
{
"role":"assistant",
"content":"<proceed_to_next_step>{\"step\": {\"step_name\": \"Publish\"}}</function>"
}
Did something change or is anyone else also facing the same? I tried with qwen3-32b and works fine, happening specifically for llama4.
Any suggestions to fix this?