mirror of
https://github.com/ByteByteGoHq/system-design-101.git
synced 2026-04-04 09:37:23 -04:00
ee4b7305a2
This PR adds all the guides from [Visual Guides](https://bytebytego.com/guides/) section on bytebytego to the repository with proper links. - [x] Markdown files for guides and categories are placed inside `data/guides` and `data/categories` - [x] Guide links in readme are auto-generated using `scripts/readme.ts`. Everytime you run the script `npm run update-readme`, it reads the categories and guides from the above mentioned folders, generate production links for guides and categories and populate the table of content in the readme. This ensures that any future guides and categories will automatically get added to the readme. - [x] Sorting inside the readme matches the actual category and guides sorting on production
1.7 KiB
1.7 KiB
title, description, image, createdAt, draft, categories, tags
| title | description | image | createdAt | draft | categories | tags | |||
|---|---|---|---|---|---|---|---|---|---|
| DeepSeek 1-Pager | Explore DeepSeek's cost-effective AI model and its innovative R1 release. | https://assets.bytebytego.com/diagrams/0164-deepseek.png | 2024-03-11 | false |
|
|
It is said to have developed a powerful AI model at a remarkably low cost, approximately $6 million for the final training run. In January 2025, it is said to have released its latest reasoning-focused model known as DeepSeek R1.
The release made it the No. 1 downloaded free app on the Apple Play Store.
Most AI models are trained using supervised fine-tuning, meaning they learn by mimicking large datasets of human-annotated examples. This method has limitations.
DeepSeek R1 overcomes these limitations by using Group Relative Policy Optimization (GRPO), a reinforcement learning technique that improves reasoning efficiency by comparing multiple possible answers within the same context.
Some facts about DeepSeek’s R1 model are as follows:
- DeepSeek-R1 uses a Mixture-of-Experts (MoE) architecture with 671 billion total parameters, activating only 37 billion parameters per task.
- It employs selective parameter activation through MoE for resource optimization.
- The model is pre-trained on 14.8 trillion tokens across 52 languages.
- DeepSeek-R1 was trained using just 2000 Nvidia GPUs. By comparison, ChatGPT-4 needed approximately 25K Nvidia GPUs over 90-100 days.
- The model is 85-90% more cost-effective than competitors.
- It excels in mathematics, coding, and reasoning tasks.
- Also, the model has been released as open-source under the MIT license.