For three weeks, we subjected Gemini Ultra 3.0 and GPT-5 to 200 identical tasks covering logical reasoning, programming, creativity, mathematics, and multimodal analysis.

Results by Category

  • Logical reasoning: GPT-5 +8 points (out of 100)
  • Python/TypeScript code: near-perfect tie
  • Creativity and storytelling: Gemini Ultra 3.0 preferred by 62% of human evaluators
  • Advanced mathematics: GPT-5 +12 points
  • Image and video analysis: Gemini Ultra 3.0 clearly superior

Which Model Should You Choose?

GPT-5 remains the champion of pure reasoning and mathematics. Gemini Ultra 3.0 excels in creativity and multimodality. For most professional uses, the two are interchangeable — the decisive factor becomes price and integration into your existing workflow.