Claude optimizes biomolecular models for speed and memory efficiency
Claude accelerated over 30 biomolecular models, improving speed and reducing memory use for larger predictions.
The optimized code for over 30 biomolecular models is open-sourced. Anthropic is also co-sponsoring a protein design competition with Adaptyv Bio, offering up to $1 million in Claude credits and wet lab validation for designs.
Scientists use open-source models for biomolecular prediction and design, but these models are computationally expensive, requiring significant resources and time.
Anthropic used Claude, specifically Claude Science, to optimize over 30 open-source biomolecular deep learning models. This involved developing custom kernels (FlashPairformer) for operations like triangle attention and multiplication, and applying model-specific optimizations such as caching redundant work. The goal was to make these models faster and more memory-efficient.
Claude sped up the models by an average of 4x with minimal precision loss, and nearly 2x with identical outputs. It also created a low-memory mode enabling predictions for systems larger than 10,000 tokens on a single GPU. This reduced the computational resources needed by two orders of magnitude compared to previous methods.
Anthropic leveraged Claude to optimize over 30 open-source biomolecular models, achieving an average 4x speedup and significantly reducing memory requirements. This makes complex biomolecular prediction and design tasks more accessible, requiring fewer computational resources.
Claude's ability to understand and modify complex code, including writing custom kernels and identifying optimization opportunities like caching, allowed for significant performance gains in computationally intensive scientific models.
- How Claude is uplifting biomolecular modeling ↗Anthropic Research · Anthropic Research · Sep 17, 2026 · paraphrased
First seen on Subvolts Sep 17, 2026. Updated Sep 18, 2026. Spotted a problem? Tell us.
- How Claude is uplifting biomolecular modelingAnthropic · Anthropic Research · Sep 17, 2026
The Playbook: reusable how-tos, prompts, configs and tips pulled out of the news.
- TemplatePrompt Library for Life Sciences TasksA collection of pre-written prompts designed for optimal results on common life sciences tasks. · 2 min
- Skill or agent configClaude Agent Skill for Single-Cell RNA QCA Claude Agent Skill to perform quality control and filtering on single-cell RNA sequencing data. · 10 min
- Cost tipReduce GPU hours for biomolecular modeling with optimized modelsUse Claude-optimized biomolecular models to achieve comparable results with significantly fewer GPU hours. · 120 min
- ToolFlashPairformer for accelerating protein structure predictionCustom kernels developed with Claude to speed up triangle attention and multiplication in protein structure prediction models. · 60 min
- ChecklistPre-computation review for Claude Science jobsCheck citations, calculations, and figure code match before submitting jobs. · 5 min
- HeuristicSecure data handling with local infrastructureRun analyses on your own infrastructure to keep large or sensitive datasets secure. · 5 min
Questions people ask
- Which AI tools does this work with?
- The sources used it with Claude.
- Where did this come from?
- Anthropic Research (Anthropic Research). Subvolts collected, normalized and credited it; the link above opens the original.