Sahaj.blogs

Day page

28 Jul 2026

Paper summary

Paper summary: Kimi K3: Open Frontier Intelligence

Kimi K3 is an open-weight 2.8T-parameter MoE (104B active) with native vision and a 1M-token context, reported as about 2.5× more scaling-efficient than Kimi K2, though still behind the strongest proprietary models.

  • 16 of 896 routed experts are active per token.
  • The report states openly that it still trails the most powerful proprietary models it evaluated.

Read the post →

Projects

Built this: Explainable Student Performance Prediction Platform

An end-to-end ML platform that validates student data, benchmarks nine model types from logistic regression to an LSTM, tracks runs in MLflow, explains predictions with SHAP and LIME, and serves them through FastAPI and Streamlit.

  • Nine models go through the same validated pipeline, from logistic regression and trees to XGBoost, CatBoost, LightGBM, PyTorch MLP/LSTM and FLAML AutoML.
  • Every prediction can be explained with SHAP and LIME.

Read the post →