在一场卡丁车游戏开发竞赛中，GPT-5.5击败了Deepseek V4 Pro。尽管Deepseek的成本仅为GPT-5.5的1/4.3，且生成了近两倍的代码量，但其最终游戏在图形、打磨和创意连贯性上均显不足。结果表明，游戏质量与成本、生成速度或代码量并无直接关联。游戏开发不仅是代码生成，更涉及系统设计、视觉判断、节奏把控和美学决策等长程规划，这些要素需协同作用才能产出优质作品。
That is a really interesting find from @atomic_chat_hq
GPT-5.5 just beat Deepseek V4 Pro in a karting game contest by turning a smaller, slower, pricier output into a better finished game.
Deepseek was 4.3x cheaper at $0.07656 and produced 18,869 tokens at 34tok/s, but its extra output did not turn into stronger graphics, polish, or creative coherence.
People still treat coding output like a volume contest, as if more tokens, lower cost, or faster generation should naturally produce a better game.
Game development punishes that assumption.
A playable karting prototype is not just code that runs. It is systems, feedback loops, visual judgment, pacing, and a hundred small aesthetic decisions that only work when they reinforce each other.
The result here shows that exactly - game quality does not track cleanly with cost, token count, or tokens per second, because game creation is not just code generation but long-range planning across physics, rendering, UI, art direction, and moment-to-moment feel.
Deepseek V4 Pro vs GPT-5.5 in a gamedev contest (full prompt is below)🏎️ Cost: Deepseek V4 Pro: $0.07656 GPT-5.5: $0.33063 Output stats: Deepseek: 34 tok/s · 9m 5s · 18,869 tokens GPT-5.5: 25 tok/s · 7m 5s · 10,580 tokens Conclusion: GPT-5.5 clearly made the better karting game. Deepseek V4 Pro was 4.3x cheaper and generated almost 2x more tokens, but the final result was weaker. It struggled with graphics, visual polish, and creative direction, while GPT-5.5 delivered better game quality, better visuals, more creativity, and stronger overall execution. Even though Deepseek positions itself as a strong model for coding, in this gamedev test it still felt far behind GPT-5.5. Try the same karting prompt with another AI model and share your result below.在 X 查看被引用的帖子
来源：Rohan Paul · x.com