[Feature] Replace gpt-3.5-turbo with gpt-4o mini

[Gonzalo-Bruna]

Motivation

OpenAI is deprecating the gpt-3.5-turbo model with the introduction of gpt-4o-mini, which is 60% cheaper and better in performance across various benchmarks.

Detailed Design According to their blogpost, GPT-4o mini surpasses GPT-3.5 Turbo in both textual and multimodal reasoning, supports the same languages as GPT-4o, and improves function calling and long-context handling (128K context window vs. 4K in GPT-3.5 Turbo).

There’s no reason to continue using gpt-3.5-turbo as it is more expensive and less capable.