kipply, Transformer Inference Arithmetic
kipply, Transformer Param Count
Jay Alammar, The Illustrated Transformer
Horace He, Making Deep Learning Go Brrrr From First Principles
Sean Goedecke, Fast LLM Inference From Scratch
DeepSeek-V3 Technical Report
Baseten, Kimi K2 Thinking at 140 tok/s on NVIDIA Blackwell
Baseten, Inference Engineering
SemiAnalysis, InferenceX v2: NVIDIA Blackwell vs AMD MI355X
swyx (Latent Space), Reasoning Price War
Cline, A Practical Guide to Hill Climbing (Evals)
How to land a job at frontier lab