Perceptic's science harness beats base model scores by up to 11.8 points on bio benchmarks
nathanbenaich · x · 2026-10-23
Perceptic released perceptic-os, a science-focused harness that scores up to 11.8 percentage points higher than the underlying model's published results on BioMysteryBench and DrugDiscoveryBench. Investor Nathan Benaich notes it makes GPT-6 astra "even more of a biobeast". The result suggests domain-specific harness engineering — not just swapping models — can substantially amplify performance on scientific tasks.
More from Models
- Leak hints a new model lands Friday at roughly 70% the cost of Kimi K3 — ChrisGPT · 2026-10-23
- Kimi teases 'Something big is coming' ahead of a Moonshot release — DotaMate · 2026-10-23
- Claude insists a cantaloupe photo is a mango in funny 'gaslighting' fail — Memoryscoop · 2026-10-23
- Pokee touts private on-prem deployments of its 10M-context model — Kyrannio · 2026-10-23
- User theory: OpenAI's constant limit resets are training Pro users before paid caps hit — Iwantthegreatest · 2026-10-23
- Kimi K3.1 rumored to be imminent as Moonshot drops a teaser signal — iamfakhrealam · 2026-10-23