# Mapping the Landscape of Large Language Models


Dominik Lukeš's lecture maps the large language models of autumn 2026. It starts from one idea, that all AI is a model plus orchestration, and covers the models; agents follow the week after. He argues that what large language models add is meaning: they take instructions in plain language and handle messy input, and are less exact, so their output has to be checked. His short version is that bigger models are better, reasoning models are better than non-reasoning ones, and semantic models are better than non-semantic ones.

## You'll come away with

- The three kinds of model, computational, predictive and semantic, and the step each does best when a customer's email is processed.
- How two training phases make a model: pre-training gives general knowledge and skills such as language, and post-training gives the chat, the instructions and the reasoning.
- How to read the model families and their generations, and how reasoning level affects score and cost, with medium to high as the sweet spot.
- Why benchmarks saturate and why some labs train for them, and what else to compare: speed, latency and price.


[Watch video](/media/2026-11-03/mapping-the-landscape-of-large-language-models-2026-11-03-summary-web.mp4) · [Captions](/media/2026-11-03/captions.vtt)

0:00 Choose your next AI model
0:48 Instructions in plain language
1:33 One customer email, three kinds of model
2:27 Post-training makes models usable
3:13 Models live in families
4:04 How much reasoning, at what cost
4:51 China: no clear winner
5:38 Model cards and saturated benchmarks
6:23 Individual benchmarks and reviewers
7:11 Reasoning: medium to high

Narration uses a synthetic voice. This video is an AI-assisted summary of the recorded lecture.

[Learning page](/learn/2026-11-03/)

[Slides](https://handouts.fyi/8st7)


[All talks](/) · [How these were made](/how/)