A 2.78-trillion-parameter model in 8 GB of RAM: a C program runs Kimi K3 anywhere — slowly
A new open-source project, kimi-k3-in-c, runs Moonshot's full 2.78-trillion-parameter Kimi K3 on a single CPU in as little as 8.24 GB of RAM — no GPU, no framework, just portable C99. It streams the 1.56 TB checkpoint from disk and keeps in memory only the 16 of 896 experts that fire per layer, producing byte-identical output whether it gets 8 GB or 224 GB. The catch is speed: about 32 seconds per token on a laptop. The repo has drawn over 2,600 GitHub stars in under a week.
Why it mattersIt won't replace anyone's chatbot, but it proves the largest open model ever released can be inspected and run by anyone with a laptop and a big SSD — turning 'open weights' from a license term into something you can actually touch. For systems programmers, it doubles as a masterclass in doing more with less.
▶ Related video: Run a Kimi K3 on 8GB RAM (No GPU)
Read in the app — free, in 9 languages
Related stories
Apple Music will tell you when a song was made by a machine — but the uploader decides whether to say so2026-08-21Stripe has just paid $7.5 billion for a model router. Days later Ramp built one and is giving it away until January.
2026-08-21Meta's assistant is now a Mac app that reads your screen and types into any window — and what it sees can train the model
2026-08-21One click on a news site now tells Google to show you more of it — and you stay on the page you were reading
2026-08-21Adobe will now generate the music, the voiceover and the door slam — and it says the licence covers you
2026-08-21