Deepseek R1 first impressions - Web + Local with Ollama

Here's my first impressions of Deepseek R1. We check out the web interface, as well as running some local models with Ollama.

https://www.deepseek.com/
https://ollama.com/library/deepseek-r1

00:00 Introduction
01:10 First Simple Prompt
01:52 Write a Poem
02:50 Write some Python Code
04:49 Translate English to Spanish
05:18 Summarize an Article
07:28 Write Hello World in Assembler
08:41 Build Hello World app from DeepSeek Code
10:25 Write a short story
11:41 Tell me what to pack for a hike
14:59 Run models locally with Ollama
15:37 Run 70B parameter model - Ollama
17:13 What should I pack - Ollama
17:34 70B Results - Ollama
18:59 Run 32B parameter model - Ollama
19:44 32B Results - Ollama
21:00 Run 1.5B parameter model - Ollama
22:25 Conclusion

DeepSeek has released DeepSeek-R1, an open-source AI model that matches the performance of OpenAI's o1 in reasoning, math, and coding tasks. This model, under an MIT license, allows for free use, modification, and commercialization, significantly reducing costs by using pure reinforcement learning. DeepSeek-R1 has shown competitive or superior results in benchmarks like MMLU, Codeforces, and GPQA. The release has sparked discussions on the implications for AI regulation, the economic impact on proprietary models, and the potential for open-source AI to advance rapidly. It's now integrated into platforms like OpenRouter and available for public evaluation, highlighting its practical applications and community engagement.

---
Follow my live streams on Twitch: https://www.twitch.tv/jeremymorgan
Follow me on YouTube: https://www.youtube.com/jeremy....morgan?sub_confirmat
Yell at me on Twitter: https://twitter.com/jeremycmorgan
Check out my Blog: https://www.jeremymorgan.com