When comparing the progress over time for each AI company, most company's models have random leaps and bounds. Anthropic is the outlier, where their model progress over time is mostly linear. https://x.com/scaling01/status/1957168153202761740
When other companies release a new flagship model that puts them at #1 on benchmarks, Anthropic tends to release a new model immediately after, regaining the throne.
These two points might suggest that Anthropic is sandbagging.
I agree with Jesper.
These two points might suggest that Anthropic is sandbagging.