1littlecoder · 2026-07-15 · notable
1littlecoder: 'NEW Tencent Hy3 is here for FREE!'
1littlecoder walks through Tencent Hunyuan Hy3, the 295B/21B-active open-weight mixture-of-experts released under Apache 2.0, and shows how to try the 256K-context model without paying an API bill.
1littlecoder shows how to run Tencent's 295B open-weight Hy3 for free.
What is it?
The July 15 video from 1littlecoder covers Tencent Hunyuan Hy3, the 295B mixture-of-experts model with 21B active parameters and a 256K context window that Tencent open-sourced under Apache 2.0 in early July. The 'free' angle is the license — the weights are on Hugging Face and can be used commercially without paying Tencent.
How does it work?
Hy3 is a sparse MoE that activates about 21 billion of its 295 billion parameters per token, so it can be served on far less GPU memory than a dense model of the same total size. 1littlecoder's video walks through hosted playgrounds and the Hugging Face weights so viewers can compare it against Z.ai's GLM 5.2 and Moonshot's Kimi K2.7 without a paid API key.
Why does it matter?
Chinese open-weight releases keep landing at the same tier as closed frontier models, and Hy3 is the second 295B-class MoE from a Chinese lab to ship this year. 1littlecoder's how-to lowers the barrier for developers who want to test Hy3 side-by-side with Claude Opus or GPT-5.6 Sol before committing to a paid stack.
Who is it for?
Developers and researchers evaluating open-weight coding and agentic models.
Try it
Watch: https://www.youtube.com/watch?v=4RyYeo5qGzU