Free Kimi K3, GLM and DeepSeek on One NVIDIA Key
Four open coding models NVIDIA hosts for free, how they score against Claude and GPT, and how to plug the key into Claude Code, Codex, Cursor and Cline.
What you get for $0
NVIDIA runs a catalog of open AI models on its own GPUs and lets developers call them for free with one API key. Four of them are strong coding models from Chinese labs, and on their makers' own benchmark tables they land right next to Claude and GPT.
| Model | Made by | Size | Coding score* |
|---|---|---|---|
| DeepSeek V4.1 Flash | DeepSeek | 552B (8B active) | 90.6 |
| Kimi K3 | Moonshot AI | about 2.8T (104B active) | 88.3 |
| GLM-5.3 | Z.ai | 753B (40B active) | 88.2 |
| GLM-5.3 Flash | Z.ai | 320B (18B active) | 84.3 |
*Terminal-Bench 2.1, a test of coding agents working in a terminal, as each maker publishes it. On the same tables Claude Opus 5 scores 89.1, Claude Fable 5 88.0 and GPT-5.6 Sol 88.8. These are self-reported, so read them as "in the same league", not as exact wins.
- No credit card and no separate subscription for each model
- One key runs all four, plus about 80 other models in NVIDIA's catalog
- It speaks the OpenAI-style API, so most coding tools can use it
- The free tier is rate-limited and meant for building and testing
Keep reading
Free, for an email.
Unlocks 5 more sections, 8 copy-paste prompts, the document version, and every other guide on the site, for free. Enter your email once to keep reading.
19,000+ follow where these guides come from.No spam. Unsubscribe in one click.