A
argbe.tech - news
1min read

OpenAI formalizes its push into research with OpenAI for Science

OpenAI is formalizing its push into scientific research with a dedicated OpenAI for Science team, as it highlights GPT-5.2’s higher performance on a PhD-level science benchmark.

MIT Technology Review reported today that OpenAI is formalizing a focused effort to support scientific work via a dedicated OpenAI for Science team.

  • OpenAI says the science-focused group was launched in October and is led by vice president Kevin Weil (previously in product leadership roles at Twitter and Instagram, and formerly a particle-physics PhD student at Stanford).
  • The company ties the push to its newer reasoning-model work, first announced in December 2024, aimed at improving multi-step math and logic problem solving.
  • On GPQA, a benchmark of 400+ multiple-choice questions spanning biology, physics, and chemistry, GPT-4 scores 39% versus an expert baseline of ~70%; OpenAI says GPT-5.2 scores 92%.
  • The GPT-5.2 update referenced in the story was released in December, and OpenAI published scientist-contributed case studies in November on how GPT-5 was used in research workflows.
  • OpenAI frames the near-term goal as speeding up research by surfacing hard-to-find prior work, helping outline proofs, and proposing lab tests, rather than claiming consistent first-discovery breakthroughs.