返回
原文
Rohan Paul@rohanpaul_ai
44AI 编辑部评分，满分 100
2026-08-01 01:47· 27分钟前
跳到正文
AI 摘要

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。

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
atomic.chat1-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 th...
推理模型发布端侧
在 X 查看原推导出 Markdown
返回精选列表
44导出 Markdown
Rohan Paul · @rohanpaul_ai · X·2026-08-01 01:47·27分钟前
在 X 看原推· x.com
AI 摘要

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。

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
atomic.chat1-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 th...
推理模型发布端侧
在 X 查看原推x.com