GPT-6 Rosalind logo

GPT-6 Rosalind

OpenAI · Released Oct 2026

Conditional

GPT-6 Rosalind is OpenAI's reasoning model for biology, drug discovery and translational medicine, built on the same weights as GPT-6 Astra and priced at $15 input and $75 output per 1M tokens at short context. The deciding caveat: access is limited to approved organizations through OpenAI's trusted-access program, so most readers cannot call it. It suits approved life sciences research teams; everyone else should use a general model. We have not tested it.

Is it right for you?

Good for

  • Approved life sciences research on biology, drug discovery and translational medicine. OpenAI's system card describes it as a frontier reasoning model for these.
  • Approved research on dual-use biological topics. OpenAI's system card says Rosalind has fewer refusals on dual-use biological queries, while still refusing biological weaponization requests.

Not good for

  • Anyone outside the trusted-access program. OpenAI says access is limited to approved customers.
  • Long-context-heavy use on a budget. The long-context rate is $30 input and $112.50 output per 1M tokens, double the short-context input rate.
  • Requests for biological weaponization. OpenAI says the model still refuses them.

How it performs by task

Medicinal chemistry

Very Good

OpenAI's own rubric: 49.6% mean score vs 28.9% for GPT-Rosalind-5.5 at the highest reasoning budget tested. Vendor figure.

Genomics and quantitative biology

Good

OpenAI's own test: 37.7% single-attempt success vs 16.7% for GPT-Rosalind-5.5. Vendor figure, and an absolute rate below 40%.

Pricing

Input

$15 / 1M tokens

Output

$75 / 1M tokens

Context

Short context shown. Long context $30 input / $112.50 output. Cached input $1.50 short, $3.00 long. Trusted-access only.

View full pricing

Benchmarks

BenchmarkScoreSource
MedChemBench v7 (99 problems), mean rubric score, highest reasoning budget (OpenAI; GPT-Rosalind-5.5 scored 28.9%)49.6% Source
GeneBench Pro v13 (129 problems), single-attempt success, highest reasoning budget (OpenAI; GPT-Rosalind-5.5 scored 16.7%)37.7% Source

No verdict changes yet

The clock starts day one. Changes land here as our verdict evolves.

Verification log

  • Pricing· No changes

    Automated agent

How we evaluate