Learning path

Open Weights Progress

The open-weights movement has reshaped what anyone — researcher, startup, or hobbyist — can run and fine-tune. This path traces the arc from the infrastructure and techniques that made open weights practical, through the labs and models that pushed the frontier, to the hardware that makes it all run. Take the steps in order: each one adds a layer to the picture.

Mixed level7 steps~42 min

7 steps

Begin →

Hugging Face
Start here: Hugging Face is the platform where open weights live and are shared — understanding it grounds everything that follows.
Read →Beginner In-depth
LoRA
The technique that made fine-tuning open models affordable — without LoRA, most practitioners couldn't adapt these models at all.
Read →Beginner In-depth
Mixture of Experts
The architecture behind many frontier open models — understanding Mixture of Experts explains how large models stay efficient enough to release.
Read →Beginner In-depth
Mistral AI
The European lab that made open weights a competitive strategy, establishing that frontier-quality models could be released publicly.
Read →Beginner In-depth
DeepSeek V4
The open-weight release that reset expectations for capability — a concrete case study in how far the frontier has moved.
Read →Beginner In-depth
NVIDIA
The hardware layer underneath it all — NVIDIA's GPUs are what actually runs these open models, and understanding the supply shapes what's possible.
Read →Beginner In-depth
Microsoft
Microsoft's role rounds out the picture: a major closed-ecosystem player that has also invested heavily in open-weight distribution and tooling.
Read →Beginner In-depth

Hugging Face

LoRA

Mixture of Experts

Mistral AI

DeepSeek V4

NVIDIA

Microsoft