Benchmark says frontier models still vary widely on antibody thermostability prediction
DeryaTR_ · x · 2026-07-25
Frontier models show uneven scientific performance
A benchmark on antibody thermostability prediction finds that newer frontier models do not always outperform their predecessors on scientific tasks.
- The shared chart ranks Opus 4.8 first with a Spearman correlation of 0.366.
- GPT 5.5 follows at 0.277, then GPT 5.6 Sol at 0.216.
- Gemini 3.1 Pro scores 0.075, and Grok 4.5 scores 0.037.
- The post argues that task-by-task benchmarking matters, especially when models are evaluated on a large set of drug-discovery workloads.
More from Research
- HandsOn benchmark: 66 verified tasks and 18,000 human demos for dexterous manipulation — stingning · 2026-10-27
- FrontierCS 2 launches on Kaggle: 48 tasks make agents beat published CS papers — MeganRisdal · 2026-10-27
- LightMTP: lightweight latent multi-token prediction boosts LLM planning in pretraining — gdm3000 · 2026-10-27
- DexPort: a handheld interface with a real robot hand for dexterous-hand data collection — pabbeel · 2026-10-27
- EMNLP study: AI literary translation reads fine but varies more, even within one book — tuvllms · 2026-10-27
- Lab monkeys at $30-80K each: a 4-year breeding cycle is the real preclinical drug bottleneck — zebird0 · 2026-10-27