Probamos GPT-5.2 vs Gemini en escenarios reales. Estos son los resultados

Probamos GPT-5.2 vs Gemini en escenarios reales. Estos son los resultados

by EDteam

Formal & Physical Sciences AI — Models & Research UYQArtificial intelligence (AI)UYQFGenerative artificial intelligence
🎙 EDteam 👥 1.0M 📅 December 17, 2025 ⏱ 39 min 👁 27K 📄 expert opinion 🧭 2026-08-02
Available in: English (current) Français

Keywords

GPT-5.2Gemini 3benchmarksproductivityAI comparison

Summary

The video compares OpenAI’s GPT-5.2 with Google’s Gemini 3, focusing on real-world performance. The presenter begins by contextualizing the release of GPT-5.2 as a response to Gemini 3, which had recently positioned Google as the AI leader. He highlights OpenAI’s ‘code red’ internal state and the rapid release of GPT-5.2. The video then examines official benchmarks, particularly GDP Val, where GPT-5.2 reportedly surpasses human experts in 70.9% of comparisons, but emphasizes the need for human supervision. It also discusses SWE-Bench Pro, where GPT-5.2 scores 55.6%, outperforming Gemini 3 on the same test. The presenter conducts hands-on tests, including an ocean wave simulation, a party card creator, and a typing game, comparing the outputs of GPT-5.2 and Gemini 3. He notes that GPT-5.2 excels in spreadsheet and presentation creation, while Gemini 3 remains superior in image generation. The video concludes that GPT-5.2 is a strong competitor but not a clear winner, with Gemini 3 still leading in certain areas. The presenter also promotes EDteam courses and discounts.

166 words

Critical Evaluation

The video provides a reasonably informative comparison of GPT-5.2 and Gemini 3, but its scientific rigor is limited. The presenter relies heavily on official benchmarks from OpenAI, which may be biased, and acknowledges this by noting that companies often fine-tune models for specific tests. However, he does not critically evaluate the benchmarks’ validity or independence. The hands-on tests are subjective and not controlled, with only a few examples, making it difficult to draw general conclusions. The presenter’s expertise is not formally established, and the video includes a promotional segment for EDteam, which could introduce bias. The analysis of the ocean wave simulation is illustrative but not systematic. The video does not cite external sources or peer-reviewed studies, and the sources listed in the description are mostly promotional links. Overall, the video offers a useful overview for a general audience but lacks the depth and rigor expected in a scientific evaluation. The adéquation titre/contenu is good, as the title accurately reflects the content. The video’s strength lies in its practical demonstrations, but its weakness is the lack of independent verification and the potential for bias due to the presenter’s affiliation with EDteam and the promotional content.

195 words

Title / Content Match

The title accurately reflects the content, which is a comparative test of GPT-5.2 and Gemini 3 in real-world scenarios.

Quality & Reliability

6/10

The video provides a comparative analysis of GPT-5.2 and Gemini 3, based on official benchmarks and hands-on tests. However, the analysis is largely subjective, lacks independent verification, and relies on promotional material from OpenAI. The presenter's expertise is not formally established, and the video includes a promotional segment.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Cited Sources

  • EDteam — Platform for programming and AI courses, mentioned in the video.
  • EDteam free courses — Link to free courses, mentioned in the video.
  • EDteam courses — Link to all courses, mentioned in the video.
  • EDteam scholarships — Link to scholarships, mentioned in the video.
  • EDteam Instagram — Social media link, mentioned in the video.
  • EDteam LinkedIn — Social media link, mentioned in the video.
  • EDteam premium — Link to premium subscription, mentioned in the video.
  • EDteam teachers — Link for teachers, mentioned in the video.
  • EDteam TikTok — Social media link, mentioned in the video.

Concurring Sources

  • OpenAI GPT-5.2 announcement — Official announcement of GPT-5.2, supporting the video's claims about its capabilities.
  • Google Gemini 3 announcement — Official announcement of Gemini 3, providing context for the comparison.

Dissenting Sources

  • Independent benchmark evaluations — The video relies on OpenAI's own benchmarks, which may be biased. Independent evaluations from third parties are not mentioned.

Contribution & Novelties

The video provides a practical, hands-on comparison of GPT-5.2 and Gemini 3, going beyond official benchmarks. It highlights GPT-5.2’s strengths in productivity tasks like spreadsheet and presentation creation, while acknowledging Gemini 3’s superiority in image generation. The presenter’s real-world tests, such as the ocean wave simulation, offer tangible insights into the models’ capabilities. However, the analysis is not exhaustive and lacks independent verification.

Pour aller plus loin :

  • GPT-5.2 official page — Official information about GPT-5.2.
  • Gemini 3 official page — Official information about Gemini 3.
  • SWE-bench — Benchmark for software engineering tasks, relevant to the programming comparison.
  • GDPval — OpenAI’s benchmark for professional tasks, mentioned in the video.

109 words

Radar Profile

The radar profile shows moderate scores across all dimensions, with quantity of information being the highest (7) and technical level the lowest (5). This suggests the video provides a good amount of information but lacks deep technical analysis, making it suitable for a general audience rather than experts.

Reliability 5/10

💬 No comments were provided for analysis.