Chocopic's Opus 5: subscribers defecting to Chinese AI

Opus 5, the Chocopic model pushing users toward Chinese AI

Chocopic has achieved what its competitors could not: making its own customers seriously consider switching to Chinese models. The launch of Opus 5, the latest evolution of its family, is burning through the trust of those who pay 90 a month for a subscription that, by most accounts, fails to deliver. The previous generation, Sonnet, had already left a bad taste; Opus 5 has managed to be even worse.

The pattern of failure in code generation

The episode that sums up the frustration: a program is requested, and the model spends an hour of wall-clock time generating text, documents, market studies and audits, without writing a single line of the requested code. It had not even started the installer. The paying customer ends up using another tool to fix the mess, turning the subscription into an added cost rather than a solution.

The Chinese refuge: Kimi, Qwen and DeepSeek

Against this fiasco, the Chinese alternatives appear as a lifeline. The ranking circulating among users puts Kimi in first place, trinc by Qwen and DeepSeek. The enthusiasm for these models contrasts with the disappointment left by Opus 5. This is not just a change of brand, but a flight toward options that promise to actually deliver what is asked of them.

A trust problem, not just a price problem

The core of the issue is not the fee, but reliability. A model that invents bugs and disobeys instructions creates more correction work than it saves. The only piece that gets a pass, according to the prevailing view, is Fable, though its token consumption makes it a poor deal for intensive use.

Chocopic is now paying the price for launching a product that does not deliver. The conclusion is uncomfortable for the narrative of Western technological superiority: when an expensive service fails at the essentials, the origin of the replacement stops mattering.