Fireship · 2026-07-22 · notable
Fireship: 'Kimi K3 just parameter mogged every open-weight model…'
Fireship's fast-cut breakdown of Moonshot's Kimi K3 — the 2.8T open-weight MoE that ranks 4th globally on Artificial Analysis and knocks every other open model down the leaderboard.
Fireship's take on Kimi K3, the 2.8T open-weight model that just moved the top of the open leaderboard.
What is it?
Fireship's July 22 video 'Kimi K3 just parameter mogged every open-weight model…' walks through Moonshot's Kimi K3 release — a 2.8-trillion-parameter mixture-of-experts model with a 1M-token context that Artificial Analysis ranks fourth overall, ahead of every other open-weight release.
How does it work?
Fireship condenses the K3 announcement into a Fireship-length explainer: how the 2.8T MoE compares to earlier Moonshot Kimi releases, where it lands against Claude Fable 5 and GPT-5.6 Sol on independent benchmarks, and what the pending July 27 weights drop means for anyone hoping to self-host it. The delivery is his usual fast-cut, meme-heavy format aimed at developers.
Why does it matter?
Kimi K3 is the biggest open-weight jump of the last quarter, and Fireship is one of the highest-reach channels developers use to catch up on releases without reading the model card. A Fireship video landing the same day as US press coverage means the K3 news is now saturated across mainstream dev audiences.
Who is it for?
developers and hobbyists who want a two-minute read on Kimi K3 without wading through the model card
Try it
Watch: https://www.youtube.com/watch?v=YP73B9D20V4