GPT-Rosalind and Cyber Models

OpenAI sells two families of specialized models, and neither is open to the public: GPT-Rosalind, "Life sciences reasoning for approved organizations", and the Daybreak cyber models, GPT-5.6 Cyber and GPT-6-Cyber, for authorized security testing. Both GPT-6 versions are built on GPT-6 Astra: GPT-Rosalind-6 uses the same weights as GPT-6 Astra, and GPT-6 Cyber is "a cyber-specialized variant of GPT-6 Astra". Access comes through OpenAI's trusted-access programs, and they run in ChatGPT Work, Codex and the OpenAI API, not the Chat tab.

Quick facts

GPT-Rosalind: life sciences

GPT-Rosalind is OpenAI's model line for biology, drug discovery and translational medicine. The research version became generally available on Sep 8, 2026, and on Oct 30, 2026 OpenAI released GPT-6 Rosalind Research for approved Research customers in Work, Codex and the API. Its system card describes it as "frontier reasoning model for biology, drug discovery, and translational medicine" and says it uses GPT-6 Astra's weights with no extra training for this release; what changes is who may use it and the safeguards around it.

Who can use it. Qualified scientists, research institutes, companies and government partners, and access is "only made available via trusted access". OpenAI reviews each applicant's intended use, security and governance; companies go through business verification and compliance screening, and approved customers must restrict access, use least-privilege controls and have qualified people review the outputs. The pricing page says the same: "Access is limited to approved internal research through a trusted-access program." For U.S. university labs, OpenAI's Rosalind Academic Plan (Oct 15, 2026) offers up to 30 Business Standard seats free for up to 12 months.

Prices, as of Nov 6, 2026 (OpenAI API pricing):

In ChatGPT workspaces on credits, GPT-Rosalind-Research costs 125 credits per million input tokens and 625 output, and the Business and Enterprise rate card lists GPT-5.5 Rosalind Research and Discovery at 20 credits a message.

What OpenAI measured. Against GPT-Rosalind-5.5, the system card reports MedChemBench 49.6% against 28.9%, GeneBench Pro 37.7% against 16.7% and LifeSciBench Gold 61.3% against 44.0%. OpenAI rates it High, not Critical, for biological and chemical capability, and it still refuses help with biological weapons.

Daybreak and the cyber models

Daybreak is OpenAI's program for cybersecurity defenders, with two tiers. The API changelog dates Daybreak Blue and Red to Aug 7, 2026 and the Codex changelog to Aug 10; Red gives "purpose-trained models such as GPT-5.6 Cyber". Using it requires OpenAI's Trusted Access for Cyber approval (learn.chatgpt.com). One privacy difference: in Daybreak sessions on a personal account, OpenAI may use your interactions to improve its models even with model training turned off.

GPT-6-Cyber (Nov 5, 2026) is for when you need to "Investigate a vulnerability, reproduce it in an authorized test environment, or verify a fix." It has Astra's 1,050,000-token context and 128,000-token output, works through Chat Completions and the Responses API (not Batch or Realtime), and requires "Daybreak Red access and approval" for the organization. Its system card adds identity and liveness checks for invited individuals, a FIDO2 security key with Advanced Account Security, SSO, MFA and API IP allowlisting (OpenAI), plus blocks on help with industrial controls, essential services and election systems.

On OpenAI's own security tests it edges past Astra: SEC-Bench Pro 87.02% against 85.38%, SRE-Bench 92.84% against 87.98%, ExploitGym 48.45% against 42.43%. The outside evaluator Irregular found it solved all 11 CyScenarioBench challenges at least once, against 8 for Astra, though on FrontierCyber it solved 90 of 288 against Astra's 96.

Prices, as of Nov 6, 2026: gpt-6-cyber $20 / $2 / $100 per 1M tokens; gpt-5.6-cyber and gpt-5.5-cyber $12.50 / $1.25 / $75. With ChatGPT credits, Daybreak Blue costs 100 / 10 / 500 credits and Daybreak Red 312.5 / 31.25 / 1,875 per million tokens. OpenAI's application security agent is a separate product; see Codex Security.

Other specialized versions

The API price list also carries gpt-6-astra-law, a legal version at $10 / $1 / $50 per 1M tokens on promotional pricing "at least through December 31, 2026". No OpenAI page read on Nov 6, 2026 describes who can use it or how it differs from Astra. For medicine, OpenAI's consumer-facing options are Health in ChatGPT and ChatGPT for Clinicians.

Frequently asked questions

What is GPT-Rosalind? GPT-Rosalind is OpenAI's life-sciences model line for biology and drug discovery, available only to approved organizations through a trusted-access program. The newest, GPT-6 Rosalind (Oct 30, 2026), uses GPT-6 Astra's weights.

Can I use GPT-6-Cyber? Only if your organization has Daybreak Red access and approval from OpenAI. Individuals must also pass identity checks and use a hardware security key. It is not in ChatGPT's Chat tab.

How much does GPT-6-Cyber cost? GPT-6-Cyber costs $20 per million input tokens, $2 cached and $100 output in the OpenAI API, as of Nov 6, 2026, twice GPT-6 Astra's Standard price.

What is OpenAI Daybreak? Daybreak is OpenAI's cybersecurity program for vetted defenders, launched in August 2026 with Blue and Red tiers. Red unlocks the cyber models, GPT-5.6 Cyber and since Nov 5, 2026 GPT-6-Cyber.

Sources

• Models, OpenAI API docs, read Nov 6, 2026
• GPT-6-Cyber model page, OpenAI API docs, read Nov 6, 2026
• Pricing, OpenAI API docs, read Nov 6, 2026
• API changelog, OpenAI, read Nov 6, 2026
• GPT-Rosalind-6 System Card Addendum, OpenAI Deployment Safety Hub, Oct 30, 2026
• GPT-6 Cyber System Card, OpenAI Deployment Safety Hub, Nov 5, 2026
• Model Release Notes, OpenAI Help Center, read Nov 6, 2026
• ChatGPT release notes, OpenAI Help Center, read Nov 6, 2026
• ChatGPT rate card (Business, Enterprise and Edu credit-based pricing), OpenAI Help Center, read Nov 6, 2026
• Data controls in ChatGPT, OpenAI Help Center, read Nov 6, 2026
• Codex pricing, OpenAI, read Nov 6, 2026
• Codex changelog, OpenAI, read Nov 6, 2026