I am using Claude for stock investment. I am automatically extracting financial statements and organizing data, but the most important thing is Claude's ability to analyze and infer data. I haven't tried GPT5.5 yet, but Opus 4.6's answer was the most satisfying. The latest version of Sonnet also seems usable. However, Claude tokens are melting too quickly, so I am wondering if it would be better to invest in a graphics card and run a local model. I wonder if there is a local model that can at least output the inference value of the latest version of Sonnet. If so, please recommend a graphics card.
