跪拜 Guibai
← Back to the summary

Free AI Models Collapse Under Daytime Load, Forcing a Retraction

Hello everyone, this is Lao Liu

A few days ago, I recommended several free models to everyone:

Lao Liu personally tested 20+ free models: only 3 are usable, and the biggest pitfall isn't the models themselves

I didn't expect to be proven wrong so quickly.

I did intensively use free models for two weeks. Based on my experience, among the many free providers, only the following are worth using for daily lightweight tasks.

My feeling at the time was that the services from these three providers were relatively stable, not prone to hitting rate limits, and I didn't notice any obvious intelligence degradation.


But I discovered I overlooked a very important issue. As free models, I mainly used them at night. During daytime working hours, I didn't dare use free models much due to security concerns.

Then yesterday, around 4 PM, I used a free model to modify a piece of my own code and suddenly realized that during peak daytime hours, these models are almost all unusable.


Setting aside security concerns, the biggest problem with free models is service stability.

So I still advise everyone not to rely entirely on free models. At least buy one subscription.

Because very often, the time you waste, the interrupted train of thought, and the frustration you feel are far more costly than the token fees you save.

🤝 If any students here are interested in client-side or Flutter development, feel free to contact Lao Liu. Let's learn from each other.

🎁 Private message to get Lao Liu's compiled "Flutter Development Handbook" for free, covering 90% of application development scenarios. It can serve as a knowledge map for learning Flutter.

💬 : laoliu_dev

📂 Lao Liu has also organized his historical articles in a GitHub repository for easy reference.

🔗 https://github.com/lzt-code/blog

Comments

Top 1 from juejin.cn, machine-translated. The original thread is authoritative.

目标艾泽拉斯

Claude bans accounts way too aggressively; better to find a way to use GPT models. I've heard the domestic model GLM 5.2 is already usable. Minimax and Kimi have inflated benchmark scores but poor real-world performance.