Alex Finn reviews Sonnet 5 as a lower-cost model for coding, tool use and agent work. He argues that its practical advantage is not replacing the strongest Opus model everywhere, but handling frequent tasks where speed and token cost matter more than maximum reasoning depth.
Alex Finn runs an informal coding comparison in which Sonnet 5 and GPT-5.5 each build a small interactive simulation. Sonnet produces a more functional result in this single prompt, while Alex Finn notes that the OpenAI model remains cheaper in some coding scenarios. The demonstration is useful directional evidence rather than a controlled benchmark.
Alex Finn's main workflow separates planning from execution. He uses a higher-compute Opus mode to ask questions, explore architecture and write a detailed plan, then switches to Sonnet 5 to implement that plan. The same cost-sensitive split can extend to persistent agents, provided users verify model identifiers, permissions and results instead of treating the cheaper model as a universal replacement.
Watch on YouTube
