Gemini 3 vs Claude Sonnet 4.5 - which model actually codes better? (Deep Dive)

Summary

In this video, the creator tests Google's Gemini 3 AI model, comparing it with Anthropic's Claude for tasks like building a landing page and creating a full-stack AI-powered to-do list app. While Gemini 3 impresses with its polished design and smooth animations for landing pages, Claude offers unique premium aesthetics. The AI manager built into the to-do app showcases Gemini 3’s ability to prioritize and organize tasks. The creator shares their excitement about Gemini 3's potential, concluding that it could be the new top model, possibly setting a new benchmark for AI tools.

Takeaways

• 😀 Gemini 3 is being tested against Anthropic's Claude in two tasks: designing a landing page and building a full-stack app.
• 😀 The landing page design test aims to assess how well Gemini 3 and Claude adhere to detailed prompts and produce visually appealing results.
• 😀 Gemini 3 produced a polished, professional landing page design with animations, a mockup of the app, and more human-like design elements compared to Claude.
• 😀 Claude's design was solid but had some visual inconsistencies, like button contrast issues, and was perceived as slightly more AI-generated compared to Gemini 3.
• 😀 In the full-stack app test, Gemini 3 generated a task manager app that prioritized tasks and allowed AI-driven suggestions, though it had some bugs.
• 😀 Gemini 3 performed better in terms of task management, with the ability to prioritize tasks and suggest next steps, offering a smoother user experience overall.
• 😀 Both Gemini 3 and Claude showed strengths in landing page design, but Gemini 3 was considered more polished, with better fonts, icons, and overall aesthetics.
• 😀 Gemini 3's full-stack app demonstrated quicker bug-fixing and smoother integration with AI, making it more useful for rapid prototyping of real-world apps.
• 😀 The test with Gemini 3 showcased its ability to build complex, AI-driven applications that manage tasks intelligently, which Claude struggled to replicate as easily.
• 😀 The main takeaway is that Gemini 3 appears to be a stronger and more polished AI model, outperforming Claude in both landing page design and app building, particularly for real-world use cases.

Q & A

What is the main purpose of testing Google's Gemini 3 in this video?

The main purpose of testing Google's Gemini 3 is to compare its capabilities with Claude and other AI models, specifically in building landing pages and full-stack applications, and to assess if it's a viable option for more complex coding tasks.

What features stood out in the Gemini 3 generated landing page design?

Gemini 3's landing page featured subtle animations, smooth hover effects, and a polished design with well-designed mockups of the app. The overall design felt more professional and less AI-generated compared to Claude.

How does Claude's landing page design compare to Gemini 3's?

Claude's landing page design was also good but felt more basic compared to Gemini 3. While it shared similarities due to the detailed prompt, Gemini 3's design was more polished, with better fonts, icons, and layout, making it feel more human-generated.

What was the key difference in the way Gemini 3 and Claude handled the prompt to improve the landing page design?

Gemini 3 made subtle improvements to the design, tweaking animations and mockups, while Claude's response was more reactive to the request, making its design feel more safe and predictable. Gemini 3, on the other hand, appeared more intuitive and polished.

What new feature was added to Gemini 3's landing page after prompting for further improvements?

After prompting for further improvements, Gemini 3 introduced a 'dark mode' design, which was highly appreciated for its visual appeal. It also added better animations, and improved the images of people in testimonials, though real people were still preferred for authenticity.

How does the user perceive Gemini 3's capabilities in full-stack application development?

Gemini 3 was tested by building a simple AI-powered to-do list app. While the app's functionality was impressive, some features, like task prioritization and task creation, needed further refinement. Overall, Gemini 3 showed potential but would need additional testing for more complex app development.

What is the role of AI in the to-do list app, and how does it manage tasks?

In the to-do list app, AI acts as a task manager, helping to prioritize tasks based on urgency and business impact. It can also add tasks, delete them, and provide suggestions for task completion, such as offering advice on what tasks to prioritize next.

What feature does the user suggest for improving the AI task manager app?

The user suggests adding a feature where the AI can send a daily email summary of tasks, offer motivational tips, and even conduct research or gather resources to assist with completing tasks.

What were some limitations encountered during the Gemini 3 task manager app testing?

Some limitations included bugs in task prioritization and difficulty in deleting multiple tasks at once. Additionally, the AI couldn't handle certain complex requests like providing gym advice when tasks were already in progress.

What is the user's overall impression of Gemini 3 after the tests?

The user is impressed with Gemini 3, praising it for being polished and capable of handling both design and coding tasks. They believe it is likely the top AI model currently available, though further testing and updates are needed for more complex applications.