NEW Google Gemini 2.5 VS DeepSeek V3.1: Who Wins? 👀

Key Concepts

Gemini 2.5 Pro Experimental, Deepseek V3.1, Reasoning Model, Base Model, Coding, Content Creation, API, Open Router, Root Code, Client, SEO ROI Calculator, AI Profit Boardroom.

Coding a Game

The video compares Google Gemini 2.5 Pro Experimental and Deepseek V3.1 by prompting them to code an endless runner game using p5js, with pixelated dinosaurs and interesting crowds. Gemini 2.5 Pro, despite being a reasoning model, returned the code faster and provided instructions. The resulting game was functional and addictive. Deepseek V3.1's game, while visually appealing, had a broken game-over trigger and inconsistent day/night cycles, making it less engaging. Result: Gemini 2.5 Pro wins in coding.

Content Creation

The models were tasked with creating SEO content for "SEO training in New York" using a prompt from the AI Profit Boardroom. Deepseek V3.1 generated HTML code, likely due to a detail in the prompt. Gemini 2.5 Pro was significantly faster. Deepseek's content was more humanized, with a compelling introduction that felt personalized and sales-oriented. Gemini 2.5 Pro's content was considered less convincing. Result: Deepseek V3.1 wins in content creation and humanization.

Example: Deepseek's opening line: "Need SEO training New York that actually works not just theory but battle tested strategies from someone who's built a seven figure agency."

Reasoning Challenge

The models were given a reasoning challenge: picking apples from a tree across a frozen river in winter. Gemini 2.5 Pro correctly identified the implausibility of finding ripe apples in winter, suggesting assessing their condition first. Deepseek V3.1 focused on crossing the frozen river without acknowledging the seasonal issue. Result: Gemini 2.5 Pro wins in reasoning and speed.

Quote: Gemini 2.5 Pro: "Assess the apples first mate are they like even worth it cuz most apples ripen in late summer and winter apples still in the tree in winter are likely frozen rotten overripe or damaged."

API Testing

The video tests the APIs of both models using Open Router keys within Visual Studio Code (Root Code) and a separate client. Initially, Gemini 2.5 Pro's API failed to work in the client. Deepseek V3.1's API worked but was slow. Later, Gemini 2.5 Pro's API started working in Root Code but proved buggy and prone to exhaustion, even for a simple SEO ROI calculator. The video notes that Deepseek's models are open source, allowing local hosting (though V3.1 wasn't available for local hosting at the time of the video).

Technical Term: Open Source - Software with source code that is available for modification and enhancement by anyone.

Result: Deepseek V3.1's API is more reliable, while Gemini 2.5 Pro's API is faster but buggy. Root Code is recommended over the client for Gemini 2.5 Pro API usage.

AI Profit Boardroom

The video promotes the AI Profit Boardroom, a community focused on using AI to increase income and save time. It offers access to prompts, tips, workflows, and courses, including a 3-hour course on Deepseek. The community provides support and weekly Q&A sessions.

Free SEO Strategy Session

A free one-to-one SEO strategy session is offered, promising to reveal how to achieve significant website traffic growth and sales increases. The session includes a custom game plan and insights into effective SEO strategies.

Synthesis/Conclusion

Gemini 2.5 Pro Experimental and Deepseek V3.1 each have strengths. Gemini 2.5 Pro excels in coding and reasoning, while Deepseek V3.1 shines in content creation with a more humanized writing style. However, Gemini 2.5 Pro's API is currently buggy and unreliable, making Deepseek V3.1 a more stable choice for API-based applications, despite its slower speed. The video emphasizes the importance of choosing the right model based on the specific task and highlights the resources available in the AI Profit Boardroom for further learning and support.