GPT-5 Academic Benchmark Chart
Academic SWE-bench Verified Software engineering Without thinking With thinking Accuracy (%) (pass @1) 74.9 52.8 69.1 30.8 GPT-5 OpenAI o3 GPT-4o Aider Polyglot Multi-language code editing Without thinking With thinking Accuracy (%) (pass @2) 88.0 26.7 79.6 26.8 GPT-5 OpenAI o3 GPT-4o
A comparison chart labeled “Academic” shows GPT-5, OpenAI o3, and GPT-4o performance on software-engineering and multilingual code-editing benchmarks. The circled scores highlight the contrast between the models, creating a reaction meme about benchmark results and the perceived gap in coding ability. EN: A comparison chart labeled “Academic” shows GPT-5, OpenAI o3, and GPT-4o performance on software-engineering and multilingual code-editing benchmarks. The circled scores highlight the contrast between the models, creating a reaction meme about benchmark results and the perceived gap in coding ability.
academic benchmark chart / академический график бенчмарковbar chart / столбчатая диаграммаpink bars / розовые столбцыwhite bars / белые столбцыcircled numbers / обведённые числаsoftware engineering label / подпись «разработка программного обеспечения»multilingual coding label / подпись «многоязычное программирование»comparison graph / сравнительный графикGPT-5 label / надпись GPT-5OpenAI o3 label / надпись OpenAI o3GPT-4o label / надпись GPT-4o
- @ithueti / 16568 2025-08-07
- Type
- image
- Size
- 800×510 · 28.4 KiB
- Mood
- sarcastic
- Language
- en
- Added