DeepSeek Vs CHATGPT: The fight for the throne!

Ok, if you are reading all the news today, it's NVIDIA and DeepSeek in a battle for the AI throne. I have been studying Deepseek for a while now and just recently DeepSeek R1 launched, and we have been given the option of installing locally, so we can test the performance without all the hype.

Ok, first, our contenders:

DeepSeek: DeepSeek was built with a focus on practicality and seamless integration into professional workflows. Its architecture is optimised for tasks requiring precision, efficiency, and domain-specific expertise. DeepSeek’s development reflects a commitment to creating an AI that empowers users in industries ranging from technology to finance, offering tools that enhance productivity and decision-making.

ChatGPT: ChatGPT, developed by OpenAI and launched in 2022, is part of the Generative Pre-trained Transformer (GPT) family, with GPT-4 being its latest iteration. It is designed to excel in natural language understanding and generation, making it a versatile tool for creative writing, brainstorming, and conversational applications. ChatGPT’s training on vast datasets allows it to engage in nuanced discussions and generate human-like text across a wide range of topics.

Here is an analogy that helps to remember:

DeepSeek: The Left Brain

• Logical and Analytical: DeepSeek thrives on structure, precision, and data-driven tasks. It’s like the left brain, which is associated with logic, reasoning, and problem-solving.
• Strengths: Optimising workflows, coding, data analysis, and technical documentation.
• Vibe: The no-nonsense, results-oriented professional who gets things done efficiently.

ChatGPT: The Right Brain

• Creative and Expressive: ChatGPT excels in creativity, imagination, and fluid communication. It’s like the right brain, which is linked to creativity, intuition, and storytelling.
• Strengths: Brainstorming, writing, generating ideas, and engaging in open-ended conversations.
• Vibe: The free-spirited, imaginative thinker who loves exploring possibilities and adding a touch of flair.

Now you cannot argue which is better as they both have widely varying applications, However, for my world which is; streamlining operations, cost reduction and efficiency this has pretty massive implications.

Let's say an insurance corporation, Busy Bee Inc, has 20 claims officers handing 50 cases per day, and want to incorporate an AI assistant to assist with call handling, claim preparation/validation and communications assistance.

Busy Bee Scenario Details (Claude/ChatGPT/DeepSeek) Locally Hosted

• Number of Claims Officers: 20
• Requests per Day per Officer: 180
• Tokens per Request: 150
• Total Tokens per Day: 3,600 requests x 150 tokens = 540,000 tokens per day

Hardware Assumptions

• GPU: NVIDIA A100 (40 GB) at $10,000 per GPU.
• CPU: AMD EPYC 7xxx (32 cores) at $3,000.
• RAM: 128 GB DDR4 at $1,000.
• Local Storage: 2 TB NVMe SSD at $500 per drive (2 drives required).
• Power Supply: 2-3 kW PSU at $500.
• Cooling System: High-performance cooling at $1,000.
• Networking: 10 Gbps network card at $500.
• Server Chassis: Rack-mounted server chassis at $1,000.

As you can see, the upfront costs, and ongoing maintenance is significant, ChatGPT is 107.14% more expensive than DeepSeek to spin up, and costs 16.72% more to run.

Mind you this is just a small scenario, so why the market is in turmoil, and NVIDIA is down 16% today, as if you want to scale this to an operation or scenario that requires hundreds of GPU's we are talking a difference of millions of dollars in operational costs.

So DeepSeek is one to watch, and I think ChatGPT has a very valid reason to be nervous.