Rohan Paul· @rohanpaul_ai · X·· 2026-08-01AI 评分44
AI 导读
Atomic Chat 将 Kimi K3 量化为 1-bit，模型从 2.8T 参数压缩至 590GB（-62%），保留 78.7% 原始质量和完整 1M 上下文。
正文 · 原文
Really cool. @atomic_chat_hq 's 1-bit Kimi K3 quant shrinks the 2.8T model to 590GB (-62%)
- The 1-bit model keeps 78.7% of the original quality and the full 1M context. model on Huggingface
- Moving to 2.1 bits raises the footprint to 769GB, but agreement jumps to 87.2%. Going all the way to 3.4 bits adds another 428GB for only 0.5 percentage points.
- In their HTML 3D physics test, all 4 models handled the drop physics, while only local K3 built the requested working winch.
- They ran it with on 4x B200 GPUs
1-bit Kimi K3 performs at Opus 5 level on 3D physics! We ran our Atomic Chat quant of Kimi K3 locally on 4x B200 against three cloud models and gave them all the same task, to build a giant anvil drop test as a single HTML file with real physics Outputs: K3 1bit (local): 15.8K tokens, $0 API cost Kimi K3 (API): 15.3K tokens, $0.30 API cost Opus 5: 22.8K tokens, $0.77 API cost GPT 5.6: 14.5K tokens, $0.72 API cost All four got the physics right. But only Kimi made a working winch. The drum turns and the chain drags the flat car off the pad. Opus 5 drew the most detail, road markings and sparks on the hit. And you can run a model at this level on your own box now. That still feels insane to us在 X 查看被引用的帖子
来源：Rohan Paul · x.com