AITOPAI热报
ProductsMulti-source

Intology 的 Locus 在分子预测上超过 GPT-6 Astra 和 Claude Opus 5.5 智能体

rohanpaul_ai@rohanpaul_ai
Post· 11 sources
View the post
Why we picked it

Intology 做了个叫 Locus 的系统，能自己提实验、并行跑、再迭代，分子预测上干过了 GPT-6 Astra 和 Claude 的智能体，做药物设计的朋友可以看看。

Intology 推出面向药物和抗体设计的 AI 系统 Locus。Locus 能自主提出实验方案，在 GPU 上并行运行多批实验，保留有效结果并迭代重复。在分子预测任务上，Locus 的成绩超过了 GPT-6 Astra 和 Claude Opus 5.5 的智能体。该系统定位是替代机器学习研究员完成分子模型设计工作。

Original · rohanpaul_ai

An AI system just designed better molecular models than the specialist tools built by the labs that own those models.

@intology’s Locus does a machine-learning researcher’s job: > proposes experiments, > runs many batches of them in parallel on GPUs > keeps what works and repeats, > now applied to drug and antibody design.

Locus beat GPT-6 Astra and Claude Opus 5.5 agents on molecular prediction

#Locus#Intology#GPT-6 Astra#Claude Opus 5.5#药物设计

Keep reading

Recent stories with the same tag or section
ProductsBuildersOfficialSelected01:32
F3 Lightning: Meta Speeds Up Ads ML Feature Deployment

Feature engineering pipelines, not model architectures, are often the real bottleneck in large-scale ads systems, and F3 Lightning targets exactly that gap. ML platform engineers building ranking or recommendation infrastructure can study how Meta compiles tens of thousands of features into one graph and what the next iteration changes. Meta published the full engineering write-up on 3 Nov 2026.

#Meta#F3 Lightning#ads ranking#ML infrastructure
官方一手Meta Engineering BlogSource
ProductsBuildersMulti-sourceSelected01:08
LlamaIndex speeds up OCRia-1 inference nearly 10× on one B200

The compounding effect of 17 small inference tweaks, rather than a single big fix, is the lesson here for anyone serving document models in production. The team's methodology of testing every change against both throughput and accuracy is reusable for other inference stacks. Contributions landing in vLLM will eventually benefit anyone running that framework directly.

#OCRia-1#LlamaIndex#vLLM#NVIDIA B200
Event
LlamaIndex@llama_index11 sourcesSource