Experience with 'Llama-3.3 70B Instruct' LLM in Oracle – Differences vs. 'GPT-5 Mini'.

We’re exploring the Oracle “free-to-use” Llama-3.3 70B Instruct model and are interested in hearing from others who have worked with it.

We currently have a seeded AI Leave Agent that works exactly as expected when using GPT-5 Mini.

However, when we switch only the LLM model—using the same agent configuration and prompts—the agent no longer behaves correctly under Llama-3.3 70B Instruct. We see that the tools are used, but no response is generated:

Since the configuration, prompt, and setup are unchanged, we’re trying to understand what might be causing the different behavior. We’re looking for insight on: