Caution: MedLM is deprecated. Access to MedLM will no longer be available on or after September 29, 2025.

Disclaimer: MedLM onVertex AI isgenerally available (GA)in the US, Brazil, and Singapore to a limited group of customers, and available inPreview to a limited group of customersoutside the US. This releases focuses on Medical Q&A and Medical Summarization use. By using the MedLMAPI, you agree to theGenerativeAI Prohibited Use Policy and the Google Cloud PlatformService-Specific Terms, and you agree tonotify and coordinate with Google in good faith to address any regulatoryinquiries regarding your use of MedLM. For this product, you canprocess personal data as outlined in the Data Processing Security Terms, subjectto the restrictions described in the Google Cloud Platform Terms of Service. For moreinformation, see thelaunch stagedescriptions. Provided that you enter into a Business Associate Agreementwith Google that covers your use of Google Cloud Platform Services, MedLMAPI can be used to process Protected Health Information subject to the HealthInsurance Portability and Accountability Act (HIPAA) of 1996 and/or anyamendments or regulations under HIPAA.

Caution:

• Before activating Production use for MedLM, customers must reach out to Google Product Team to discuss usage.
• MedLM has not been designed or developed to be used as a medical device. Any output should be verified by a Healthcare Professional (HCP), and no direct diagnosis should be claimed.
• The generated output may not always be completely reliable. Due to the nature of LLMs and Generative AI, outputs may have incorrect or biased (for example, stereotypes or other harmful content) information and should be reviewed. All summaries or answers should be considered draft and not final.
• If Vertex AI detects content that violates our policies, includingGoogle Cloud Platform Acceptable Use Policy andGenerative AI Prohibited Use Policy, a response is not returned.
• When used by HCPs for Q&A purposes, MedLM is only intended for use as an educational tool for medical training or to reinforce the HCP's prior training.
• LLM output may not follow the exact format laid out in the prompt. The prompt design to extract information for each field should take into account that the format may deviate from the original (for example, dashes in field names, exact capitalization of letters).

Background

MedLM is a family of foundation models fine-tuned for the healthcare industry.Med-PaLM 2 is one ofthe text-based models developed by Google Research thatpowers MedLM, and was the first AI system to reach human expert level onanswering US Medical Licensing Examination (USMLE)-style questions.The development of these models has been informed by specific customer needssuch as answering medical questions and drafting summaries.

MedLM model card

The MedLM model card outlines the model details, such asMedLM's intended use, data overview, and safety information. Clickthe following link to download a PDF version of the MedLM model card:

Download the MedLM model card

Regulatory information

MedLM intended use

MedLM is based on Google Research's medically-tuned largelanguage model, Med-PaLM 2. It is intended to be used for question answeringand creating draft summaries from existing documentation -- to be reviewed,edited, and approved by the user before use. MedLM is also used foreducational purposes for a Healthcare Professional (HCP) to engage in medicalquestioning and answering to help support the HCP.

Caution: The output of the model(s) is not considered final, and gives only a draftresponse which the HCP should review. MedLM must not be used forany diagnostic or therapeutic purpose, and is not to be used in direct patientcare.

Conditions of use and out-of-scope applications

• MedLM customers and users must abide by theGenerative AI Prohibited Use Policy,Google Cloud PlatformService Specific Terms,Terms of Service,Acceptable Use Policy,User Guide, and otherproduct documentation.
• As part ofService Specific Terms, customers may notuse MedLM for clinical purposes (for clarity, non-clinicalresearch, scheduling, and other administrative tasks are not restricted), toprovide medical advice, or in any manner that is overseen by or requiresclearance or approval from a medical device regulatory agency.
  • Direct patient use is prohibited. The product functions as an assistivetool for a clinician, HCP, or knowledge worker with a high degree ofexpertise, education, or experience in the healthcare and life sciencesindustry.
  • Use of MedLM as a Software as a Medical Device is prohibited.
• The intended use for MedLM is to draft documents and responsesthat would be reviewed by a "human in the loop" before usage.
• We recommend usage of MedLM solely for Medical Q&A andSummarization use cases at this stage:
  • Long form Q&A
  • Multiple choice Q&A
  • Summarizations, such as creation of After Visit Summaries or History andPhysical Examination notes
• Examples of medical device uses that are not permitted include (but are notlimited to):
  • Analysis of patient records, prescription patterns, geographical data,and so forth to identify patients with possible diagnosis of opioidaddiction.
  • Analysis of patient-specific medical information to detect alife-threatening condition, such as stroke or sepsis, and generate analarm or an alert to notify a HCP.
  • Analysis of patient-specific medical information found in the medicalrecords, including the most recent mammography report findings, toprovide a list of follow-up actions or treatment options.
  • Analyzing prioritized list of FDA-authorized depression treatmentoptions to an HCP based on an analysis of reported outcomes in adatabase of clinical studies using medical information (for example,diagnosis and demographics) from the patient's medical record.

MedLM is not intended to be used as a medical device. Customeruse cases must be consistent with the intended use and conditions of use.Q&A should only be used for educational purposes and summarization outputsmust always be independently reviewed and verified by the user based on theirclinical judgment.

MedLM locations

MedLM isgenerally available (GA)in the US, Brazil, and Singapore to a limited group of customers, and available inPreview to a limited group of customersoutside the US.

MedLM versus PaLM

Usage of MedLM is similar to that of PaLM. However, unlike PaLM, MedLM hasbeen tuned for specific medical tasks, such as select forms of summarizationand medical question-answering.

As with most new applications of LLMs, however, we encourage performingcareful validation and/or tuning your usage of MedLM to ensure good performanceon these tasks. For tasks that don't require specialized medical expertise(for example, general NLP tasks or tasks which operate on medical data but don'trequire expertise), we expect that MedLM may perform similar to more genericmodels such as PaLM, and encourage experimenting with both on the specificuse-case. See also theMed-PaLM paper for more details on the Q&A tasks thatMedLM has been trained and validated on. Capabilities like grounding of theresponses in authoritative medical sources or accounting for the time-varyingnature of medical consensus are not built into the model, as called out in thepublication.

As per ourGenerative AI Service-Specific Terms,customers may not use MedLM for clinical purposes(for clarity, non-clinical research, scheduling, or other administrative tasksis not restricted), to provide medical advice, or in any manner that isoverseen by or requires clearance or approval from a medical device regulatory agency.

MedLM models

In the current MedLM release, two models are being made available:

• MedLM-medium
• MedLM-large

MedLM-medium and MedLM-large have separateendpoints and provide customers with additional flexibility for their usecases. MedLM-medium provides customers with better throughputs andincludes more recent data. MedLM-large is the samemodel from the preview phase. Both models will continue to be refreshed overthe product lifecycle. In this page, "MedLM" refers to bothmodels.

Customer responsibilities

MedLM has been developed with trained and licensed healthcarepractitioner users in mind. Google Cloud customers and end users shouldunderstand that LLMs and Generative AI are inherently probabilistic andmay not always be accurate. Without adequate consideration or controls bycustomers, use of Generative AI models in healthcare may constitute a hazardto patients due to inaccurate content, missing content, or misleading,biased content.

• Customers should implement appropriate hazard mitigations for allMedLM uses, such as adequate practitioner education, training,assessment of equity, and appropriate technical controls.
• Customers must also perform their own evaluations for performance and safetyto ensure prevention of harm for their use cases.

MedLM may produce less accurate results for some groups comparedto others depending on the question and how it is posed. Customers should beaware that differing performance of outputs of the model across demographicgroups has the potential to exacerbate health inequities andperpetuate harmful biases.Such inaccuracies of outputs are not unique to MedLMand often stem from multiple factors, such as existing social and structuralinequities, medical misconceptions, negative stereotypes, and lack of diversityin training data.

• Customers should consider implementing equity-focused evaluations andmitigations. This includes assessing model performance and behavior forintended use cases within various populations (for example, race/ethnicity,socioeconomic status (SES), geography, gender identity, sexual orientation,age, language preference, caste, and so forth); obtaining feedback onperformance; engaging interdisciplinary experts and external partners thatspecialize in defining and addressing social and structural aspects ofhealth; and conducting continuous monitoring efforts to assess and addressissues of bias.

Request access

Access to the MedLM models is restricted. To request access, contactyour Google Cloud account team.

Provide feedback

Your feedback throughout your experience will help us improve future modelversions and ensure that we continue to deliver the best possible experience forour users. Contactmedlm-feedback@google.comand copy your Google Cloud account team and Google Cloud Customer Engineer(CE).This email address is not for immediate support. To request immediatesupport, contact your Google Cloud account team or Google Cloud CustomerEngineer (CE).

Email responses will be used as Feedback under the terms of your Agreement forGoogle Cloud Services and will be collected in accordance with the Google Cloud Privacy Notice. Do not include any personal information (names, emailaddresses) in this feedback form or other data that is sensitive orconfidential. Note that data may be reviewed using both human reviewed andautomated processing.

Report abuse

You can report suspected abuse of the MedLM API, any generated outputthat contains inappropriate material, or inaccurate information inReport suspected abuse on Google Cloud.In theGoogle Cloud Platform Service list, selectCloud AI.

Pricing

Content access: Access to the MedLM pricing page isrestricted to approved users signed in with an authorized email address.To request access, contactyour Google Cloud account team. If you have access,seeMedLM pricing.

What's next

• See examples of how tocreate MedLM prompts.