Insilico's 2.6B Liquid AI-based model sets SOTA in single-step retrosynthesis on 46M reactions
JosephJacks_ · x · 2026-09-04
Insilico Medicine put a 2.6B-parameter Liquid AI language model into its MMAI Gym and reports the resulting insilicoSOTAFM model now sets state of the art in single-step retrosynthesis. Trained on 46M reactions, it offers diverse predictions only partially intersecting the reaction space of previous small SOTA models, a notable AI4Science result.
Related event: Insilico's 2.6B Model Sets New Retrosynthesis SOTA(3 posts)→
More from Models
- ARC-AGI-3 is now saturated, prompting calls for new benchmarks ASAP — kimmonismus · 2026-09-04
- Astra model card: no CoT-monitor evasion when reasoning must be verbalized — bookwormengr · 2026-09-04
- OpenAI Researcher roon: GPT-6 Astra Will Be Obsolete in Weeks — Tolopono · 2026-09-04
- Reasoning effort switching without breaking cache is live in Codex and Claude — altryne · 2026-09-04
- Mollick: GPT-6 autonomously does meaningful work for days, built an open-source Alexandria Library simulation — emollick · 2026-09-04
- OpenAI's GPT-6 Astra system card: first model to hit Critical cybersecurity capability — kaicathyc · 2026-09-04