The plan is likely to bundle access to GPT-Rosalind, OpenAI's first domain-specific frontier reasoning model for life sciences research, which was upgraded to run on the GPT-5.5 agentic coding and tool-use architecture on June 3, 2026 . Currently, GPT-Rosalind is gated behind a U.S.-only enterprise offering and a global invite-only research preview . If launched, ChatGPT for Science could standardize access for a much broader set of academic and research institutions .

What the leaked code reveals

The code references place ChatGPT for Science alongside existing vertical subscription tiers for universities, finance, and government . The stated objectives in the code include "accelerating scientific progress" and "getting institutions invited to science" .

Key details from the leak:

• The plan is labeled ChatGPT for Science and targets scientific and academic institutions .
• It is designed to grant access to specialized AI tools for research, likely including or bundling GPT-Rosalind capabilities .
• OpenAI has not announced the plan publicly; sources estimate an announcement could be weeks away based on the code's freshness .

What GPT-Rosalind can do

GPT-Rosalind is described by OpenAI as a "frontier reasoning model built to support research across biology, drug discovery, and translational medicine" . Its capabilities fall into several categories:

Biological reasoning

• Deep expertise in molecular biology, protein engineering, medicinal chemistry, and biological data analysis .
• Optimized for long-horizon quantitative biology analyses, reportedly using ~31% fewer tokens than GPT-5.5 on sustained biological reasoning tasks .
• Can synthesize evidence across scientific literature, internal databases, and specialist models .

Drug discovery

• Supports early-stage drug discovery workflows, including molecular property prediction, hit identification, and lead optimization .
• Integrated with Codex plugins for life sciences, including a Lab Data Viewer and a Literature Search plugin that connect to real-world lab instruments and scientific databases .
• Already used by pharmaceutical partners such as Novo Nordisk for real R&D workflows .

Genomics analysis

• The June 2026 update deepened performance in genomics compute, allowing the model to process and reason across large-scale genomic datasets .
• Supports variant interpretation, gene-disease association analysis, and multi-omics data integration .

Agentic tool use

• Can orchestrate workflows across scientific tools, literature APIs, internal lab systems, and specialist models autonomously .
• OpenAI introduced a dedicated evaluation benchmark called LifeSciBench to measure performance on real-world life sciences tasks . On expert-judged benchmarks, GPT-Rosalind on GPT-5.5 scored 27.5% versus GPT-5.5's 25.1% on MedChemBench; 21.6% versus 20.4% on GeneBench while using 31% fewer tokens; and 63.2% versus 55.8% on LabWorkBench .

How ChatGPT for Science would broaden access

Currently, GPT-Rosalind has two narrow access paths:

• Enterprise-only for U.S.-based enterprise customers via ChatGPT, Codex, and API .
• Invite-only trusted-access research preview for select pharmaceutical and research institutions globally (expanded worldwide on June 3, 2026) .

The ChatGPT for Science plan would change this by:

• Offering a dedicated subscription tier that any eligible scientific or academic institution could purchase, rather than requiring custom enterprise negotiations .
• Sitting alongside existing vertical plans (ChatGPT for University, for Finance, for Government), suggesting a standardized product with clear pricing and features .
• Potentially broadening access to GPT-Rosalind's capabilities to university labs, academic research groups, and smaller research institutions that cannot afford enterprise contracts or do not qualify for the invite-only preview .

Caveats and uncertainty

• No official announcement: OpenAI has not confirmed the plan. All evidence comes from code references and third-party reporting .
• Pricing unknown: No pricing details have been spotted in the code .
• Eligibility unclear: It is unknown whether the plan will be open to all researchers or restricted to accredited institutions .
• Timeline speculative: Reports say an announcement could be "weeks away," but that is an estimate based on the code being recent rather than a confirmed date .

The plan is likely to bundle access to GPT-Rosalind, OpenAI's first domain-specific frontier reasoning model for life sciences research, which was upgraded to run on the GPT-5.5 agentic coding and tool-use architecture on June 3, 2026 . Currently, GPT-Rosalind is gated behind a U.S.-only enterprise offering and a global invite-only research preview . If launched, ChatGPT for Science could standardize access for a much broader set of academic and research institutions .

What the leaked code reveals

The code references place ChatGPT for Science alongside existing vertical subscription tiers for universities, finance, and government . The stated objectives in the code include "accelerating scientific progress" and "getting institutions invited to science" .

Key details from the leak:

• The plan is labeled ChatGPT for Science and targets scientific and academic institutions .
• It is designed to grant access to specialized AI tools for research, likely including or bundling GPT-Rosalind capabilities .
• OpenAI has not announced the plan publicly; sources estimate an announcement could be weeks away based on the code's freshness .

What GPT-Rosalind can do

GPT-Rosalind is described by OpenAI as a "frontier reasoning model built to support research across biology, drug discovery, and translational medicine" . Its capabilities fall into several categories:

Biological reasoning

• Deep expertise in molecular biology, protein engineering, medicinal chemistry, and biological data analysis .
• Optimized for long-horizon quantitative biology analyses, reportedly using ~31% fewer tokens than GPT-5.5 on sustained biological reasoning tasks .
• Can synthesize evidence across scientific literature, internal databases, and specialist models .

Drug discovery

• Supports early-stage drug discovery workflows, including molecular property prediction, hit identification, and lead optimization .
• Integrated with Codex plugins for life sciences, including a Lab Data Viewer and a Literature Search plugin that connect to real-world lab instruments and scientific databases .
• Already used by pharmaceutical partners such as Novo Nordisk for real R&D workflows .

Genomics analysis

• The June 2026 update deepened performance in genomics compute, allowing the model to process and reason across large-scale genomic datasets .
• Supports variant interpretation, gene-disease association analysis, and multi-omics data integration .

Agentic tool use

• Can orchestrate workflows across scientific tools, literature APIs, internal lab systems, and specialist models autonomously .
• OpenAI introduced a dedicated evaluation benchmark called LifeSciBench to measure performance on real-world life sciences tasks . On expert-judged benchmarks, GPT-Rosalind on GPT-5.5 scored 27.5% versus GPT-5.5's 25.1% on MedChemBench; 21.6% versus 20.4% on GeneBench while using 31% fewer tokens; and 63.2% versus 55.8% on LabWorkBench .

How ChatGPT for Science would broaden access

Currently, GPT-Rosalind has two narrow access paths:

• Enterprise-only for U.S.-based enterprise customers via ChatGPT, Codex, and API .
• Invite-only trusted-access research preview for select pharmaceutical and research institutions globally (expanded worldwide on June 3, 2026) .

The ChatGPT for Science plan would change this by:

• Offering a dedicated subscription tier that any eligible scientific or academic institution could purchase, rather than requiring custom enterprise negotiations .
• Sitting alongside existing vertical plans (ChatGPT for University, for Finance, for Government), suggesting a standardized product with clear pricing and features .
• Potentially broadening access to GPT-Rosalind's capabilities to university labs, academic research groups, and smaller research institutions that cannot afford enterprise contracts or do not qualify for the invite-only preview .

Caveats and uncertainty

• No official announcement: OpenAI has not confirmed the plan. All evidence comes from code references and third-party reporting .
• Pricing unknown: No pricing details have been spotted in the code .
• Eligibility unclear: It is unknown whether the plan will be open to all researchers or restricted to accredited institutions .
• Timeline speculative: Reports say an announcement could be "weeks away," but that is an estimate based on the code being recent rather than a confirmed date .