Alex: why he won't be using Sonnet 5.5

1:59 clip · English with English subtitles · Moonshots EP 299

Alex Wissner-Gross calls Sonnet 5.5 a strange release: on a cost-per-attempt basis it scores lower than Opus 5.5, while a smaller, distilled model normally delivers more intelligence per dollar. He acknowledges the big jump over the previous Sonnet and leaves room for token, latency or other considerations.

Source: Peter H. Diamandis, “Recursive's $670M Bet on Self-Improving AI, Sonnet 5.5 Hits 70%, Elon Co-Leads Pentagon Push” (EP 299), segment 1:44:21–1:46:19. Watch on YouTube · Download MP4