OpenAI has opened applications for a new program that will give 100,000 academic researchers free access to its most powerful AI tools, including GPT-5.6 Sol Pro and the Codex coding platform. The first 10,000 spots will be filled this summer, with the full rollout planned through 2027. For scientists, mathematicians, and engineers at qualifying universities, this represents a substantial—and cost-free—upgrade to their research workflows, provided they can navigate the application process and institutional policies.
What the Free Access Covers
The program bundles three main tools: ChatGPT (the standard chatbot interface), ChatGPT Work (designed for extended projects like grant writing and literature reviews), and Codex (for coding, data analysis, and building reproducible workflows). At launch, participants will get GPT-5.6 Sol Pro, part of the GPT-5.6 model family that also includes the faster GPT-5.6 Luna and everyday GPT-5.6 Terra.
What distinguishes this from the free tier available to anyone is the scale and specificity. OpenAI says approved researchers receive higher usage limits, larger context windows—crucial for digesting lengthy papers and datasets—and expanded Deep Research capabilities that can autonomously browse and synthesize information. An annual subscription would otherwise cost roughly $200 per month for similar access, according to reporting from Axios.
The toolset is heavily geared toward the life sciences. Over 75 dedicated skills cover genetics, genomics, sequencing, single-cell analysis, protein modeling, and drug discovery. There are also connectors to pull in data from scientific literature databases, public genomic and clinical repositories, satellite imagery, computational notebooks, and reference managers. In practice, this means a researcher could ask ChatGPT to scan GeneBank, summarize a selection of papers, draft a methods section, and then have Codex run a statistical analysis on their dataset—all within a single, privacy-protected workspace.
On the privacy front, OpenAI states that data entered by program participants is not used to train its models by default, and the workspaces have business-grade security. That’s a critical assurance for university IT departments and research compliance officers, though it doesn’t override local rules for handling regulated, proprietary, or export-controlled data.
Who Qualifies and How to Apply
Eligibility is restricted to researchers at recognized, degree-granting colleges and universities with a high level of research activity. Applicants must verify their institutional affiliation and describe their active research and how they intend to use the AI tools. The process is not an open freebie; OpenAI is curating participants based on scientific merit and need.
Approved researchers can invite up to four collaborators from the same institution, but each person counts toward the 100,000 total. That cap may fill quickly if large labs onboard entire teams, so early application is wise. Institutions that already use ChatGPT Edu—a separate education package—will have their program access coordinated through the existing workspace, potentially streamlining the approval process for individuals there.
Applications are open now, and the first cohort of 10,000 researchers will gain access this summer. Initial rollouts are confirmed at the Institute for Advanced Study in Princeton and France’s École normale supérieure. Others can expect to hear back on a rolling basis.
What This Means for Windows-Based Research Teams
At first glance, the program is platform-agnostic: all tools are accessible through a web browser, making them usable on any Windows PC without special software. Codex, in particular, can be integrated into existing Windows development environments like Visual Studio Code or accessed via a terminal, which will appeal to computational researchers who already rely on Windows for their data pipelines.
For IT administrators at universities, the news raises both opportunities and compliance questions. Because data is not used for training by default and the environment is business-grade, the program aligns better with institutional privacy standards than the consumer version of ChatGPT. However, Microsoft’s enterprise agreements and education licensing often include built-in AI tools through Windows Copilot and Azure OpenAI Service. IT shops will need to decide whether to direct researchers toward this OpenAI-branded program, integrate it with existing Microsoft offerings, or manage it as a shadow IT risk. At minimum, they should communicate the availability of the program to their research community and clarify data handling policies—especially for clinical, patented, or export-controlled work.
On the researcher side, Windows users may find that Codex’s ability to script and automate tasks fits neatly into PowerShell or WSL (Windows Subsystem for Linux) workflows. For example, a researcher could ask Codex to write a Python script that scrapes a public database, then run it locally in WSL, all while keeping subsequent analysis within the secure ChatGPT workspace. The larger context windows also mean that loading the full text of several papers at once—a common pain point in current free AI tools—becomes feasible.
How OpenAI Got Here: The Scientific Computing Push
This academic program is the latest step in OpenAI’s broader ambition to embed its models into serious research. The company has committed more than $250 million through 2027 to external scientific research and discovery, following initiatives like NextGenAI and a collaboration with the U.S. Department of Energy’s Genesis Mission, which provided Codex access to national lab and university researchers.
The company says that roughly 1.3 million people already use ChatGPT weekly for advanced science and mathematics, generating about 8.4 million messages. The new program aims to convert that grassroots adoption into institutionally supported, structured workflows—complete with training, support, and a feedback channel to the company.
OpenAI also points to internal benchmarks to justify the program’s model choices. GPT-5.6 Sol scored 83% on FrontierMath Tier 4, a test of research-level mathematical reasoning, compared with 72.5% for its predecessor GPT-5.5. On GeneBench Pro, which measures complex biological data analysis, GPT-5.6 Sol Pro achieved 31.5%. These numbers are vendor-reported, not independently verified, but they suggest meaningful gains in domains that academics care about.
Still, a significant limitation for many AI researchers is that the program does not provide access to model weights or training data. As Axios noted, that leaves unanswered the need for independent inspection, reproducibility, and safety evaluation—a sticking point for those who see open science as central to AI development. For researchers who just need a powerful tool to accelerate their existing work, this may not matter. But those who plan to publish findings that depend on the model’s internal behavior may face pushback from peer reviewers who insist on full methodological transparency.
Practical Steps for Researchers and IT Departments
If you’re a Windows-using researcher, here’s how to move forward:
- Check your eligibility. Confirm that your institution is a recognized, degree-granting university with high research activity. If you’re unsure, check with your office of research or library.
- Prepare your application. Clearly outline your active research project and how you plan to use ChatGPT, ChatGPT Work, or Codex. Emphasize concrete tasks: generating hypotheses, analyzing large datasets, automating data cleaning, drafting manuscripts, or preparing grant proposals.
- Coordinate with IT. If your university has an existing agreement for ChatGPT Edu or Microsoft AI tools, talk to your IT department. They may be able to integrate your access with existing identity management and data protection protocols, avoiding duplicated accounts and compliance headaches.
- Once approved, configure your workspace. Ensure the data usage setting is set to not train models (it should be default, but verify). If you plan to invite collaborators, do so early—remember that each person counts against the total program headcount.
- Be mindful of data you upload. Even with business-grade protections, don’t upload protected health information (PHI), classified data, or proprietary industry data unless you have explicit clearance from your institution’s legal or compliance office. The tool is powerful, but it’s not a magical safe harbor from FERPA, HIPAA, or ITAR.
For IT administrators, the immediate tasks are communication and policy. Send a bulletin to research groups about the program and clarify which data types are permitted. If you already manage Azure OpenAI services, compare the feature sets—OpenAI’s program includes life-science-specific skills and connectors that may not be in the standard enterprise offering. Decide whether to endorse, monitor, or restrict use.
Outlook: The First Wave of a Growing Trend
OpenAI’s move signals that frontier AI access is becoming a new kind of research infrastructure, akin to supercomputing grants or journal subscriptions. The initial 10,000-user deployment will test whether the tools meaningfully accelerate scientific output or simply add another layer of automation that requires constant supervision. Early feedback from the Institute for Advanced Study and École normale supérieure will be telling.
The program’s success hinges on more than just model capability. Universities must develop clear guidelines for attribution, integrity, and human validation of AI-assisted work. If those pieces fall into place, we could see a rapid expansion of similar programs—from both OpenAI and its competitors—locking in academic users early. For Windows-using researchers and the IT teams who support them, the message is clear: the tools are here, the access is free, but the responsibility for using them wisely still rests with you.