Tops GPTs 
LifeSciBench is an expert-authored benchmark of 750 tasks that tests AI on realistic life-science workflows. OpenAI reports detailed rubric scoring and independent validation, and shows GPT‑Rosalind improves over GPT‑5.5 but important gaps remain in artifact- and design-heavy tasks.
OpenAI reports that GPT‑5.4, paired with Molecule.one’s Maria, proposed TEMPO as an additive that improved Chan–Lam coupling for primary sulfonamides. In Maria Lab’s 10,080 reactions, mean yield rose from 16.6% to 25.2% and bench validation confirmed gains for 11 of 14 pairs.
Researchers used OpenAI o3 Deep Research to reexamine 376 previously unsolved rare-disease cases and produced evidence-linked hypotheses that led to 18 confirmed diagnoses (4.8%). The study, published June 18, 2026 in NEJM AI, emphasizes expert review and CLIA confirmation.
OpenAI says GPT‑5.5 Instant, released May 2026, narrows the gap with frontier models on health tasks and is now available to free ChatGPT users. Evaluations and physician reviews show gains in urgency recognition, context-seeking, and clearer explanations of uncertainty.
On June 18, 2026 OpenAI introduced credit usage analytics and updated spend controls for ChatGPT Enterprise. Admins can now view ChatGPT and Codex credit consumption in the Global Admin Console and set workspace, group, and individual limits.
Samsung Electronics will provide ChatGPT Enterprise and Codex to all employees in Korea and to all Device eXperience (DX) staff worldwide, the companies announced. The tools will be used across R&D, manufacturing, marketing and corporate functions, with Codex showing strong recent user growth in Kor
Daybreak’s Patch the Planet pairs AI-assisted research with human security review to find and help patch vulnerabilities in critical open-source projects. The initial surge, announced June 22, 2026, has involved 19 projects and collaboration with Trail of Bits, HackerOne, and Calif.
Omio says it has connected OpenAI models to more than 3,000 transportation providers across 47 countries and cut product development effort to about 20% of previous levels. CTO Tomas Vocetka says the company moved from search-based booking to conversational travel using ChatGPT and Codex.
Immunologist Derya Unutmaz used GPT‑5 Pro to reanalyze a 2022 experiment and identified a mechanistic link: deoxyglucose appears to disrupt IL-2 production, favoring Th17 development. The model also correctly simulated a CD8+ lymphoma killing result.
OpenAI and Broadcom on June 24, 2026 presented Jalapeño, an accelerator built for LLM inference and co-developed with Broadcom and Celestica. OpenAI says engineering samples run GPT‑5.3‑Codex‑Spark at target power, with early tests showing higher performance per watt.