# Anthropic says Claude speeds up 30+ biomolecular models by average 4x

Digest AI · Research · published 2026-09-19T04:26:00Z

Canonical: https://digestai.news/story/anthropic-says-claude-speeds-up-30-biomolecular-models-by-average-4x

## Summary

Anthropic reported that Claude, operating within the Claude Science workbench, optimized more than 30 open‑source AI models used for biomolecular tasks in just under four weeks. The reported speedup averages roughly 4× when a small precision loss is tolerated, and nearly 2× when output must match the original exactly. Two members of Anthropic’s technical staff with biomolecular modeling expertise supervised the experiment, though they had no prior experience in inference optimization or kernel engineering.

The optimization covered models for structure prediction, protein design, genomics, and protein language modeling, including AlphaFold3, OpenFold3 and Boltz‑2. Claude also introduced a low‑memory “Big” mode that enables inference on sequences larger than 70,000 tokens using a single NVIDIA GPU node, a task previously out of reach. Anthropic open‑sourced 36 optimization kits on GitHub under the Apache‑2.0 license, providing modes such as exact, fast, and big for each model. The company notes that the results are an internal case study and have not been independently verified, and the repository is marked as reference‑only and not maintained.

## Key points

- Claude optimized over 30 open‑source life‑science models in under four weeks, achieving about 4× average speedup (≈2× with identical output).
- The work was supervised by two Anthropic staff with biomolecular expertise but no prior inference‑optimization experience.
- Anthropic released 36 optimization kits on GitHub under Apache‑2.0, enabling low‑memory “Big” mode for >70,000‑token inference on a single GPU.

## Why it matters

Shows generative AI can accelerate scientific modeling, lowering compute costs and speeding research, though results are internal and not independently validated.

## Sources

1. [Claude Science optimizes over 30 AI models in under 4 weeks—averaging 4x speedup](https://note.com/aitrendlab/n/n1d9c7d5ce7ba?hl=en) (note.com, 2026-09-19)

Part of the developing story: [AI Boosts Biomolecular Modeling Speed](https://digestai.news/thread/nvidia-bioir-boosts-boltz-2-protein-folding-throughput-by-2-9x-on-h100-gpus) (2 stories)

## Cite

Digest AI, "Anthropic says Claude speeds up 30+ biomolecular models by average 4x", 19 September 2026, https://digestai.news/story/anthropic-says-claude-speeds-up-30-biomolecular-models-by-average-4x

---

Written by Digest AI's editorial model from the linked sources; the sources are the record. Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse
JSON: https://digestai.news/story/anthropic-says-claude-speeds-up-30-biomolecular-models-by-average-4x.json