Claude, Gemini, Grok & ChatGPT: My AI Subscription Stack Analyzed

The author subscribes to four major AI services — Claude Pro, Gemini Pro, Grok, and ChatGPT Plus — to maximize productivity for both primary work and side projects. Each service is evaluated on model performance, unique features, cost-effectiveness, and integration into daily workflows.

Claude Pro

Claude’s model capability is described as top-tier, especially Opus 5.5, which leads in almost every aspect. The author highlights its video generation ability (implemented via code rather than native multimodality), a previously weak link in their workflow. Despite calling the company “annoying,” the model is deemed indispensable. The author expresses hope that domestic Chinese models (DeepSeek, Kimi, GLM, Qwen) will become more competitive.

ChatGPT Plus & Codex

Although the author considered canceling, they renewed due to “withdrawal symptoms” and the superior mobile/desktop app experience. GPT-series models are seen as trailing Claude, and domestic models have further reduced reliance on Codex for coding. The Plus quota is largely symbolic; the author exploits quota resets to run short, intensive tasks with the best available model (referred to as GPT-6). A key advantage is that the web version calculates quota independently, enabling scheduled tasks for large-scale information gathering. The decisive retention factor is GPT Image 2, considered the best text-to-image model; all the author’s WeChat article covers, including this one, are generated with it.

Grok

The Grok subscription was initially purchased for the “full suite” (Cursor, Grok, Grok Bot, X), but the suite proved disappointing. The only component actively used is Grok Bot; the author has built six to seven bots that automate various tasks. A dedicated article on this setup is planned.

Gemini Pro Annual

This was a one-time $199 annual purchase. Early access to Claude models on the Antigravity platform was enjoyable, but the experience later degraded (“turned into a mess”). The release of Gemini 3.8 Flash reversed the sentiment: it is fast, has extremely generous quotas, and excels at multimodal tasks and writing. The author uses it for nearly all project proposals. The annual bundle includes an unexpected bonus — Colab Pro with free access to an 80 GB NVIDIA A100 GPU, which the author used to deploy Qwen3.8-27B locally, single-handedly justifying the $199 cost.

Motivation and Outlook

The driving force behind maintaining multiple subscriptions is the need to dramatically increase work efficiency, freeing time for side projects (some public, some private). The article ends with an open question: can the subscription fees be earned back? A follow-up piece covering domestic Chinese AI product subscriptions is promised.