Skip to main content

100+ Frontier LLMs for $0: NVIDIA Support Lands in DevoxxGenie

· 4 min read
Stephan Janssen
Stephan Janssen
Creator of DevoxxGenie

Here's a quietly wild fact: NVIDIA is casually giving you access to a whole catalogue of frontier AI models - including five of the strongest Chinese frontier models - for free. 😳

No credit card. No subscriptions. Just one API key that unlocks everything.

Starting with DevoxxGenie 1.8.11, the plugin ships NVIDIA as a first-class cloud provider. Drop in your nvapi-... key, and the entire NVIDIA build catalogue shows up in the model dropdown - over 100 models, all reachable through a single OpenAI-compatible endpoint.

Selecting an NVIDIA model in DevoxxGenie

What you get for $0​

A single free key unlocks a deep bench of frontier models. The headliners:

  • DeepSeek V4 Flash - ultra-fast inference for quick iterations
  • MiniMax M3 - a drop-in coding companion for your daily work
  • Qwen3.5-397B - heavyweight reasoning for the hard problems
  • Kimi K2.6 - agentic workflows and long tool chains
  • GLM 5.1 - a reliable, do-everything everyday model

And that's just the top of the list. The provider fetches the full live catalogue straight from NVIDIA, so you get Llama, Nemotron, Mistral, GPT-OSS, embeddings, and more - 100+ models in total, without us hardcoding a stale list.

Why this is a big deal​

  • No juggling subscriptions for different models - one key covers them all
  • No workflow changes - everything is OpenAI-compatible, so it just plugs in
  • No vendor lock-in - swap models freely inside the same DevoxxGenie session

Around 40 requests per minute is plenty for most developers and personal projects, and five of these models genuinely compete with GPT- and Claude-class output - for a grand total of zero dollars.

Getting started takes under 2 minutes​

  1. Go to build.nvidia.com
  2. Sign up and verify your account
  3. Generate your nvapi key
  4. In DevoxxGenie, open Settings → LLM Providers, pick NVIDIA, and paste your key
  5. Choose any model in the dropdown and start building

Entering your NVIDIA API key in DevoxxGenie settings

That's it. Under the hood, DevoxxGenie points at NVIDIA's OpenAI-compatible base URL - https://integrate.api.nvidia.com/v1 - and reuses the same battle-tested client path as our other cloud providers. If you'd rather wire it up manually, the base URL and your nvapi key are all you need.

Type-to-filter, because 100+ models is a lot​

A catalogue this big would drown a normal dropdown. So the NVIDIA model picker is type-to-filter: start typing kimi, deepseek, or qwen and the list narrows instantly to matching models. There's also a Refresh button that re-probes NVIDIA's catalogue on demand, so newly-released models show up without a plugin update.

Because NVIDIA's /v1/models endpoint only reports model IDs (no context length), DevoxxGenie applies a best-effort context-window heuristic per model family - 128K for most instruct models, larger for Llama 4, smaller for embeddings and safety guards. That value only drives the token-usage bar and the "add project to context" budget; it never truncates your conversation. And if the endpoint is ever unreachable, the provider degrades gracefully to a curated fallback list so you're never left without models.

A quick field guide​

A pragmatic way to split the work across the free tier:

  • DeepSeek V4 Flash when you want speed
  • Qwen3.5-397B when the reasoning is genuinely hard
  • Kimi K2.6 for agent mode and long tool chains
  • MiniMax M3 as your everyday coding companion
  • GLM 5.1 as the reliable default

Claim it before the limits change​

Five frontier models that go toe-to-toe with GPT and Claude, plus a hundred more - all behind one free key, all inside your IntelliJ IDE. Grab your key at build.nvidia.com, select NVIDIA in DevoxxGenie, and start coding. 👀

DevoxxGenie is open source on GitHub - issues, ideas, and stars all welcome.