Science & Tech · AI · 2 hrs ago
SK Biopharm builds AI model for drug development
SK Biopharm has completed testing an AI model for drug development as part of a South Korean government-led project.
The company worked with Lunit, Kakao Healthcare and other firms and medical institutions to develop AI models for medicine and science.
Its drug research model was trained on 7.14 million records, covering 66 tasks such as predicting drug toxicity and clinical trial results.
The company says testing, including checks with pharmaceutical data that is not public, showed the model could be useful for drug research.
It is also exploring AI for clinical development, including patient outcome analysis and selecting and designing clinical trial participants.
Drug development can take years, and the company aims to use AI to improve its speed and chances of success.
SK Biopharm plans to expand AI use from early research through clinical trials, using computing infrastructure based on Nvidia B200 GPUs.
SK Biopharm said it completed validation work on an AI foundation model project focused on drug development.
The project consortium developed a medical-science AI foundation model and specialized models for drug development, omics and clinical work.
SK Biopharm built 7.14 million pieces of specialized training data, amounting to 4.4 billion tokens.
The company said it tested the drug-development model on multiple tasks and used pharmaceutical companies’ confidential drug data for further validation.
SK Biopharm plans to expand AI use across drug development, from identifying drug candidates through clinical trials.
- Who
- SK Biopharm and a consortium that included Lunit, Trillion Labs, Aizen Science, Kakao Healthcare and medical institutions.
- What
- They developed and tested AI foundation and specialized models for medical science, including drug development.
- When
- SK Biopharm announced the completion of the project’s validation work on November 6, 2026.
- Where
- Not stated.
- Why
- To advance AI use in drug development and explore applications in pharmaceutical research and clinical development.
This story does not have two clearly opposing sides.
This story does not have a timeline yet.
- Training data
- 7.14 million pieces, or 4.4 billion tokens
- Drug-development tasks
- 66, including drug properties and toxicity, drug-target interactions, drug response and clinical trial outcome prediction
- Model areas
- Drug development, medical science omics and clinical work
- Computing infrastructure
- NVIDIA GPU B200-based AI computing infrastructure
- Project lead
- South Korea’s Ministry of Science and ICT oversaw the project
1 source
What people are saying
The competitiveness of AI for drug development depends on how meaningful an answer it can provide in research settings.
We will use AI to increase the speed of drug development and the likelihood of success.
We also explored the potential for AI use in clinical development, including pharmaceutical R&D work, patient prognosis analysis, and the selection and design of clinical trial participants.