Some AIs have issues with agent

[mymel2001-holder]

Issue Type

Agent Issue

Operating System

Linux

Description of the bug

Some LLMs have issues with agent, even if they should be agentic. Some examples but is not limited to:

• "deepseek-v3.1" via Ollama Cloud (Ollama and OpenAI compatible endpoints. Able to do one round of thinking before breaking.)
• "openai" via Pollinations (OpenAI Compatible endpoint, Does not function)
• Any of these broken LLMs via OpenWebUI's OpenAI compatible endpoint (such as the ones above)

Steps to Reproduce

1. Open OS
2. Go to agent
3. Add a broken LLM that shouldn't be broken (such as the ones above)
4. Try to do a task via the agent

Screenshots / Videos

No response

BrowserOS Version

137.0.7215.69

Additional Context

I am not on GitHub much, and my inbox is always pretty full, so I likely won't reply.