Recently, due to sluggish progress from Google and OpenAI in releasing new models, Claude Opus 4.6 has held the number one spot on the AI API usage charts (as the most popular model) for quite some time. However, its developer, Anthropic, is facing widespread public suspicion that it is suffering from a lack of computing power, leading to frequent "dumbdowns" and processing errors recently.
Opus 4.6 is the latest model released by Anthropic. Its API pricing is astronomically high ($5 per million input tokens, $25 per million output tokens), and it performs exceptionally well in programming, even capable of completing small projects independently. But recently, a large number of developers and power users have reported that it is getting increasingly difficult for the model to follow instructions, and it's even making amateur mistakes. Anthropic is being accused of secretly downgrading the model's compute allocation, making its output noticeably worse than before.
In the intense competition among the Big Three, OpenAI is desperately expanding its compute capacity, and Google has amassed a massive amount of computing power leveraging its years of cloud infrastructure. Anthropic, on the other hand, hasn't invested as heavily in expanding its hardware, which has also led to their recent imposition of peak-hour usage limits.
This silent nerf has triggered a massive wave of dissatisfaction among users. I have to say, the notorious "A-chu" (a sarcastic Chinese community nickname for Anthropic) really lives up to its reputation.
Comments NOTHING