When TechRadar put Gemini 3.0 Pro, ChatGPT 5.1, and Claude Sonnet 4.5 to the test on a real coding assignment, it wasn’t just another benchmark — the results revealed clear strengths and critical differences. Here’s a breakdown of why Gemini 3.0 came out ahead, especially for developers building web-based applications.
1. The Test Project: “Thumb Wars” Game
TechRadar’s author asked each model to build a web-based prototype of a game called “Thumb Wars,” where users control cartoon thumbs in a wrestling ring-like setting. The prompt was moderately detailed, leaving room for creative coding decisions (HTML, CSS, JS, controls, UI).
-
Gemini 3 Pro immediately understood the concept, suggested building a Progressive Web App (PWA), and provided robust HTML and CSS to simulate 3D-style ring depth.
-
It even added keyboard controls (for desktop) without being explicitly asked — showing reasoning about usability.
-
Over iterations, Gemini improved the visuals: adding perspective effects, more realistic thumb shapes, camera shake for “heavy hits,” and depth layering.
The result: a playable prototype, built quickly, that closely matched the author’s vision — with minimal manual corrections needed.
2. How ChatGPT 5.1 Fared
ChatGPT 5.1 also handled the task but with some notable limitations:
-
It took longer to produce its initial HTML/CSS/JS code.
-
The first version lacked desktop-friendly controls, meaning keyboard or trackpad input wasn't integrated by default.
-
When prompted to improve, ChatGPT added more realistic ring and thumb visuals, but still fell short of Gemini’s depth and interactivity. The second version of the game felt more static and “less alive.”
-
Because of this, the author felt the game lacked “vibe” — it was functional, but not immersive.
3. What About Claude Sonnet 4.5?
Claude Sonnet 4.5 demonstrated solid enthusiasm, but struggled to match Gemini’s execution:
-
It generated a prototype with character customization (skin tone, masks), game area, and basic combat mechanics.
-
However, its implementation of desktop (keyboard) controls was missing, despite repeated prompting.
-
Unlike Gemini, which reasoned about 3D movement (z-axis) and layered visuals, Claude’s version remained quite flat and had more limited motion logic.
-
Overall, the author concluded that Claude 4.5 simply didn't fill in the gaps as intuitively as Gemini.
4. Why Gemini 3.0 Stood Out: Key Strengths
a) Intuitive Reasoning & Understanding
Gemini didn’t just respond to the prompt — it interpreted intent. Its ability to deduce what a usable game would need (like keyboard controls) without explicit instructions revealed a deeper understanding of user experience.
b) Creative Iteration
The model didn’t just deliver a barebones game. Its iterative improvements — CSS perspective, realistic thumb shapes, more “dramatic” visual effects — showed that Gemini can evolve design over multiple prompt rounds.
c) Speed + Usability
Gemini coded quickly, and the resulting prototype was usable almost right away. That’s a big plus for developers who want to prototype fast or build MVPs with AI assistance.
d) “Smart Defaults”
Rather than waiting for constant prompting, Gemini made good default choices (like adding keyboard controls) — a sign that it’s been trained to think practically, not just syntactically.
5. Limitations & Trade-offs
Gemini’s performance was strong, but not without caveats:
-
Some design choices (like ring visuals) weren’t perfect — the author noted a few odd stylistic decisions.
-
The game was still relatively simple; Gemini couldn’t generate a full backend server or multiplayer real-time logic out of the box.
-
Because it “fills in” a lot of gaps, power users might prefer more control (via other models) when they want to write very custom or production-grade code.
Also, not every user agrees entirely — some community developers have reported mixed experiences in other environments, suggesting Gemini’s agentic planning isn't always perfect.
6. Comparing with Claude Sonnet 4.5: Different Strengths
While Gemini excelled in this specific “vibe-coding” scenario, Claude 4.5 maintains strengths in other areas:
-
Claude’s Artifact system (used in past tests) is great for structured prompts and code generation when you want explicit, guided feedback.
-
Claude tends to follow prompt instructions more conservatively, which can be useful when stability matters.
-
In very detailed or enterprise-grade projects, Claude may provide more predictable, step-by-step execution rather than creative “filling-in.”
7. Verdict: When Gemini 3.0 Is the Better Choice
Based on the TechRadar test, Gemini 3.0 Pro is a top pick when:
-
You want rapid prototyping for web-based apps.
-
Your workflow benefits from creative iteration — not just rigid instruction-following.
-
You appreciate an AI that can reason through UX needs, not just generate code blindly.
-
You’re experimenting or building MVPs and want smart defaults (keyboard controls, layout, 3D perspective) without micromanaging every detail.
On the other hand, if you're focused on production-level code or need very controlled, stable builds (especially for multi-user apps), models like Claude still make good sense.
Final Thoughts
The TechRadar experiment highlighted a major leap: Gemini 3.0 Pro isn’t just an AI coder — it’s a reasoning partner. In building the “Thumb Wars” game, it didn’t just churn out code — it understood player experience, UI logic, and control mechanisms, turning a rough idea into a playable web app.
For developers and creators, that shift matters. As generative AI becomes more deeply integrated into design and prototyping workflows, a model that can think ahead — not just write — may be the real edge.
