@UnTalNixon_exe: Forget about GPUs and million-dollar clusters. They just made the world's largest open model (Kimi K3 – 2.78 trillion p…
Summary
A new tool called kimi-k3-in-c runs the 2.78T-parameter Kimi K3 open model on a single CPU with as little as 8.24 GB RAM, streaming experts from disk and achieving deterministic output at 10-32 seconds per token.
View Cached Full Text
Cached at: 08/05/26, 06:27 PM
Forget about GPUs and million-dollar clusters.
They just made the world’s largest open model (Kimi K3 – 2.78 trillion parameters) run on a single CPU with just 8.24 GB of RAM.
It’s called kimi-k3-in-c.
→ Full engine: 176 KB of pure C99
→ Zero frameworks, zero BLAS, zero CUDA, zero GPU
→ The 1.56 TB checkpoint lives on disk
→ Only 16 of 896 experts activate per token
→ the rest streams on-demand
→ Memory is a dial: 8 GB, 32 GB, 128 GB or 224 GB
→ same bit-identical output In 8 GB it takes ~32 s/token. With more RAM it drops to ~10 s/token.
It’s not fast for chat… but it runs. And it does so deterministically and verified. This is the closest we’ve come to having an AI supercomputer on your laptop.
Repo
Has anyone run it yet? What preset did you use?
Similar Articles
Run Kimi K3 using 29 GB of RAM at 0.50 tok/s
WASTE is a new open-source C inference engine that streams expert weights from disk to run the 2.78-trillion-parameter Kimi K3 model on a consumer laptop with just 29 GB of RAM, achieving 0.49–0.54 tokens/s.
@Saboo_Shubham_: Someone just ran Kimi k3 with 2.78 trillion parameter on a single CPU in 8.24 GB of RAM. 100% Opensource.
Someone reportedly ran the open-source Kimi k3 model with 2.78 trillion parameters on a single CPU using only 8.24 GB of RAM, showcasing extreme efficiency.
Kimi K3 full model running on 16x GB10 cluster at 20+tps
Kimi K3 full model runs on a 16x GB10 cluster at 20+ tokens per second average, with plans to publish the vllm image and instructions.
Kimi K3 (2.8T) at 1 token/s on a MacBook Pro, streamed from four SSDs
Deltafin is an open-source tool that enables running the full Kimi K3 2.8T parameter model on consumer hardware like MacBook Pro by streaming experts from SSDs, achieving performance around 1 token per second.
Kimi K3 in and on C (c99 and cpu)
A project claims to run Kimi K3, a 2.78T parameter MoE model, on CPU with 8GB RAM by streaming expert weights from NVMe SSD and using MXFP4 compression, trading speed for memory efficiency.