Brother Long recommends a fun and free tool. If you want to do things with large model APIs, the most annoying part is spending money and the entry barriers!

OpenAI requires linking a credit card, Claude needs an overseas phone number, and domestically, there are only a few models like Tongyi and Wenxin, plus you have to register multiple accounts and manage numerous keys—purely a hassle!
Until I discovered NVIDIA's Build platform, which is simply legendary—91 free models available to use freely, including DeepSeek, Kimi, GLM, Llama—all in one, one key to access everything!
Key points: no credit card required, no total usage limits, speed limited to 40 requests per minute, supports +86 phone number verification, fully compatible with API and OpenAI, so the code I wrote before almost needs no changes!
A 5-minute registration process, so simple even a fool can do it:
1. Open the site
2. Verify email, enter the 6-digit code;
3. Create a cloud account (only English letters/numbers), verify with a +86 phone number (if you don’t receive it, switch to a different mobile number);
4. Generate an API key, select permanent validity, copy and save (only shown once!).
Here are some good free models I recommend, just copy the homework:
- Kimi K2.5: supports Chinese, multimodal, the first choice for daily creation;
- DeepSeek-R1: excellent at logical reasoning and programming, transparent thought process;
- Nemotron-3: around 1 million context length, unbeatable for long text processing.
Calling is super simple, just fill in the base_url.
A quick reminder: free tier is limited to 40 requests per minute, popular models may get congested during peak times, so using it off-peak is more enjoyable.