OpenAI prices GPT-Rosalind-6 at $15/$75, 1.5x Astra

OpenAI's release of GPT-Rosalind-6 on October 30 marks a shift in how the company packages safety for specialized domains. The model is described in the system card addendum as 'a frontier reasoning model for biology, drug discovery, and translational medicine,' yet it ships with the same weights as GPT-6 Astra. The product difference is policy: fewer refusals on dual-use biology topics, balanced by institutional vetting and qualified human review before outputs inform experiments or decisions.

The pricing tells a similar story. At $15/$75 per million tokens, GPT-Rosalind-6 costs 1.5 times Astra's $10/$50, a premium that appears to monetize the governance gate rather than any underlying capability gain. On internal benchmarks, the jump from GPT-Rosalind-5.5 is substantial—MedChemBench v7 rose from 28.9% to 49.6%—but these gains are attributed to the base Astra upgrade, not biology-specific fine-tuning. The addendum rates the model High in Biological and Chemical domains, below Critical, and confirms that none of the four Critical capability evaluations met the indicative threshold.

The launch arrives amid a dense run of OpenAI safety announcements, including restrictions on teen weapons information and auto-ended violence chats in ChatGPT a day earlier. Those moves and this one point in opposite directions on where OpenAI draws the safety line for different audiences—tightening consumer protections while loosening research access under strict vetting.