DeepSeek is here to unseat OpenAI from its top position as the leader in AI. Ever since its launch earlier this month, the Chinese AI lab has been hogging all the limelight. In simple words, DeepSeek is doing what Sam Altman-led OpenAI initially intended to do. DeepSeek has open-sourced a model that has been trained with large-scale reinforcement learning, and that too at a fraction of the cost of its bigger competitors, such as Meta and OpenAI. DeepSeek has also released a paper detailing its entire process. A few days ago, DeepSeek released its latest AI model, DeepSeek-R1, which is a state-of-the-art reasoning model designed to enhance reasoning and analytical abilities. The new model comes in two versions: DeepSeek-R1 and DeepSeek-R1-Zero. Ever since the launch of the new model, the AI community has been abuzz with comparisons between DeepSeek’s AI models and other frontier models from OpenAI, Meta, and Microsoft. Several users have taken to X to share their findings with the new DeepSeek-R1 and how it stands out in the world of AI models. Here is a look at some notable posts about DeepSeek-R1. Runs perfectly on smartphones Paul Couvert, whose X bio describes him as an AI and tech educator, shared a remarkable aspect of the DeepSeek-R1. According to his post, DeepSeek R1 version 1.5B runs smoothly on his smartphone. “So you can have a model that outperforms GPT-4o and Claude 3.5 Sonnet on math in your pocket,” he wrote. Wow DeepSeek R1 version 1.5B runs perfectly locally on my phone 😳 So you can have a model that outperforms GPT-4o and Claude 3.5 Sonnet on math in your pocket. Mind-blowing pic.twitter.com/0J3uLXI6ZR — Paul Couvert (@itsPaulAi) January 22, 2025 Rotating triangle with a red ball Another user, ZeroEdge, shared a post comparing OpenAI o1 Pro and DeepSeek R1. The user asked the models to implement a rotating triangle with a red ball in it. He used the prompt, "Write a Python script for a bouncing red ball within a triangle to make sure to handle collision detection properly. Make the triangle slowly rotate. implement it in Python. Make sure the ball stays within the triangle." ChatGPT o1 Pro vs. DeepSeek R1 Asked to implement a rotating triangle with red ball in it. Left OpenAI right DeepSeek pic.twitter.com/KB2CCkE3Fi — zeroedge (@nobody_qwert) January 21, 2025 Chat with PDF file Sumanth, another user, showed how DeepSeek R1 is the best open-source reasoning model that is on par with OpenAI’s o1. He built a RAG app to chat PDF files using the DeepSeek R1 model that can run locally on your computer. DeepSeek released R1, the best open-source reasoning model that is on par with OpenAI's o1! I have built a RAG app to chat with your PDF files using the DeepSeek R1 model, running locally on your computer. 100% Open Source pic.twitter.com/8fDpI6ReYH — Sumanth (@Sumanth_077) January 21, 2025 AGI at Home Developing artificial general intelligence at home would be a stretch. However, X user Alex Cheema shared a post where he is seen running DeepSeek R1 across seven M4 Pro Mac minis and an M4 Max MacBook Pro with 496GB unified memory. He did this by leveraging EXO Labs’ distributed inference and 4-bit quantisation. He said that the next challenge would be achieving fp8 precision, which would require over 700GB unified memory for even greater performance. AGI at home Running DeepSeek R1 across my 7 M4 Pro Mac Minis and 1 M4 Max MacBook Pro. Total unified memory = 496GB. Uses @exolabs distributed inference with 4-bit quantization. Next goal is fp8 (requires >700GB) pic.twitter.com/Sqn3doiBEG — Alex Cheema - e/acc (@alexocheema) January 21, 2025 Run locally on browser X user Vaibhav Srivastav shared that he was able to run the DeepSeek R1 1.5B version successfully on the web browser locally. In his post, he said that he did this at 60 tokens per second, backed by WebGPU. Let’s fucking goo!! DeepSeek R1 1.5B running FULLY LOCALLY in your browser at 60 tok/ sec powered by WebGPU🔥 Intelligence truly is too cheap to meter! ⚡️ pic.twitter.com/P3BOSDeFwQ — Vaibhav (VB) Srivastav (@reach_vb) January 21, 2025 Building Perplexity Clone This user created a clone of the AI-powered conversational search engine Perplexity AI. Riley Brown, used DeepSeek AI model to create the engine in just an hour, and he asserted that he did not write a single line of code. I built Perplexity but with @deepseek_ai Reasoner model in about an hour and i'm posting the full video of it later today. As always, didn't write a single line of code. Should I open source this one? pic.twitter.com/tSxMhQNTwu — Riley Brown (@rileybrown_ai) January 21, 2025 How to access DeepSeek-R1 One can explore DeepSeek-R1 through the DeepSeek Chat interface by visiting chat.deepseek.com. A working email would be needed to sign up and by clicking on 'DeepThink' on the page can let you access it. Those who want to access API should get their API key from the DeepSeek Developer Portal, set up their environment with tools like Python requests or OpenAI package, and configure your API client with the base URL - api.deepseek.com. In less than a week since the launch of DeepSeek-R1, the AI model created by the Chinese AI lab has triggered a barrage of reactions in the AI community.
