Google DeepMind Releases Free 'How to Scale Your Model' Book, Adds New GPU Thinking Chapter
mdancho84 · x · 2026-09-26
Google DeepMind's free online book How to Scale Your Model — by Jacob Austin, Sholto Douglas, Roy Frostig, and colleagues — systematically demystifies how LLMs scale on TPUs and GPUs, now updated through Part 12: "How to Think About GPUs."
The book covers:
- Hardware fundamentals: Roofline models, how TPU/GPU devices work and communicate
- How LLMs actually run on real hardware
- Parallelism strategies for training and inference at massive scale
- Practical estimation: how much an LLM should cost to train, how much memory you need to serve a model yourself, and concepts like AllGather
The authors aim to replace the "alchemy" framing with simple principles that apply from a single accelerator to tens of thousands of devices. Readers need basic LLM/Transformer knowledge and ideally some JAX familiarity.
Related event: DeepMind Releases Free LLM Scaling Guide With New GPU Chapter(2 posts)→
More from Infra
- Why this builder quit server racks: fried motherboards and a ~$1,500 housing bill — TheZachMueller · 2026-09-26
- Stanford and NVIDIA's CLM embeds decisions instead of tokens, 13x faster but accuracy drops at scale — Prompt Engineering · 2026-09-26
- Farmer's photo exposes 62 unpermitted gas generators powering Microsoft AI data center, $1.1M fine — mkheck · 2026-09-26
- LLM routing saved 33.2% vs premium models in 640-request pilot, but a fixed mid-priced model beat it — smakosh · 2026-09-26
- Akamai CEO on $11.6B Anthropic cloud deal: 'this business is going to help our margins' — pdamodaran · 2026-09-26
- MLXUI: Open-Source Local AI Browser for Apple Silicon with One-Click Model Installs — WebAssemblyMan · 2026-09-26