Pick a model, we'll tell you what to buy
These are big open-source models pulled from the world agent leaderboard, filtered to models with a genuinely open license (MIT or Apache 2.0) and roughly 100 billion parameters or more.
What's a "parameter"? Think of parameters as the number of tiny dials a model can adjust to represent what it knows.
A model named like
qwen3.5-397b has 397 billion of them. More parameters generally means smarter, but also means more memory to hold them all.
Apache 2.0 · Open Source
Qwen3.5-397B-A17B
A mixture-of-experts model: 397 billion parameters live in memory, but only 17 billion "activate" per response — which is why it runs faster than its size suggests.
≥ 476 GBminimum memory needed
MIT · Open Source
GLM-5.2 (Max)
Currently the highest-ranked MIT-licensed model on the public agent leaderboard — built for multi-step, tool-using AI agents, not just chat.
≥ 904 GBminimum memory needed
MIT · Open Source
DeepSeek-V4-Pro
The biggest model in our store: 1.6 trillion parameters. Even though it only "actively thinks" with 49 billion of them per word, every one of the 1.6 trillion has to be held in memory somewhere.
≥ 1,920 GBminimum memory needed