Free account = 1 chapter of every course unlocked
No credit card · Google sign-in in 30 seconds · 20+ free chapters across 25 courses
🎯
Pick the Right Model for Your Machine
Stop guessing which model to run. Verdict-first picks for 8, 12, 16 and 24GB+ cards and for each job you actually do, then the reasoning underneath: the three numbers that decide everything, quant formats, when a 7B beats a 70B, real usable context, sampling settings, honest speed measurement, how to read a leaderboard without being fooled, and how to build an eval from your own work.
14 chaptersFirst chapter free to preview
Full syllabus
2
Pick by Job: Coding, Writing, Vision, Embeddings, Tool-Calling
3
The Three Numbers That Decide Everything
4
Quant Formats Decoded: GGUF, AWQ, GPTQ, MLX
5
When a 7B Beats a 70B: Task Fit Over Size
6
Context Window Claims vs Real Usable Context
7
Sampling Settings: The Free Quality Upgrade Everyone Skips
8
Measuring Speed Honestly: tok/s and Time to First Token
9
How to Read a Leaderboard Without Being Fooled
10
Contamination, Cherry-Picking, and Vendor Benchmarks
11
Build Your Own Eval: 20 Prompts From Your Real Work
12
Licences Before You Ship
13
Keeping a Live Shortlist as Models Churn
14
Capstone: Your Personal Model Bench
Unlock all 14 chapters
Plus 24 other courses — 547 more chapters included.