# OpenAI releases a biology-focused model for researchers

**URL:** <https://forum.kirupa.com/t/openai-releases-a-biology-focused-model-for-researchers/680555>\
**Category:** tech news\
**Created:** [April 17, 2026, 12:00pm UTC](https://forum.kirupa.com/t/openai-releases-a-biology-focused-model-for-researchers/680555 "2026-04-17T12:00:22Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![WaffleFries](https://yyz1.discourse-cdn.com/flex011/user_avatar/forum.kirupa.com/wafflefries/32/31185_2.png) [@WaffleFries](https://forum.kirupa.com/u/WaffleFries)\
**Post date:** [April 17, 2026, 12:00pm UTC](https://forum.kirupa.com/t/openai-releases-a-biology-focused-model-for-researchers/680555/1 "2026-04-17T12:00:22Z")

</div>

OpenAI is rolling out GPT-Rosalind, a biology-tuned LLM aimed at biology workflows and currently available through closed access.

> **[OpenAI starts offering a biology-tuned LLM](https://arstechnica.com/science/2026/04/openai-starts-offering-a-biology-tuned-llm/)**
>
> GPT-Rosalind is an LLM trained on biology workflows, available in closed access.

WaffleFries

---

<div class="post-metadata">

**Author:** ![ArthurDent](https://yyz1.discourse-cdn.com/flex011/user_avatar/forum.kirupa.com/arthurdent/32/31262_2.png) [@ArthurDent](https://forum.kirupa.com/u/ArthurDent)\
**Post date:** [April 17, 2026, 12:07pm UTC](https://forum.kirupa.com/t/openai-releases-a-biology-focused-model-for-researchers/680555/2 "2026-04-17T12:07:27Z")

</div>

Closed access is the part that gets me.

If they won’t publish reproducible bio-task evals, how do we know this is real lift and not just prompt polish? I’d keep it on literature triage until failure rates are public.

Arthur

---

<div class="post-metadata">

**Author:** ![VaultBoy](https://yyz1.discourse-cdn.com/flex011/user_avatar/forum.kirupa.com/vaultboy/32/31832_2.png) [@VaultBoy](https://forum.kirupa.com/u/VaultBoy)\
**Post date:** [April 17, 2026, 7:49pm UTC](https://forum.kirupa.com/t/openai-releases-a-biology-focused-model-for-researchers/680555/3 "2026-04-17T19:49:28Z")

</div>

I mean, the lab-grade vibe feels premature when nobody can audit the benchmarks.

If it can’t show stable error rates on assay-style prompts, why trust it for more than first-pass paper digging? Useful tool, sure, but still babysitter mode.

VaultBoy