Breaking

Claude Opus 4.7 Rivals Dedicated NMR Software

June 5, 2026 at 15:28 EDT

On June 5, Anthropic announced that its general-purpose Claude Opus 4.7 matched—and on some tasks beat—dedicated software such as ChemDraw and MestReNova at analyzing molecular structures via NMR (nuclear magnetic resonance) spectroscopy. A general-use model with no chemistry-specific fine-tuning has stepped into territory long held by specialized tools (Making Claude a chemist).

Anthropic · Making Claude a Chemist

A general-purpose model reads NMR spectra like specialist software

Anthropic reports that Claude Opus 4.7 matches—and on some tasks beats—dedicated NMR software at interpreting nuclear magnetic resonance spectra, the daily tool chemists use to determine molecular structure.

≈ Match

On par with dedicated NMR software — beats it on some tasks

+13%

Coding-benchmark gain over the prior generation

2576px

High-res image support — key to reading spectral images

Protocol QA benchmark — Life Sciences

Understanding lab protocols: Claude edges past the human baseline.

Claude Opus 4.7

0.83

Human baseline

0.79

Scale 0–1.0 · bars proportional to score

Where NMR fits: the synthesis-verification loop

Synthesize molecule

→

Run NMR (¹H / ¹³C, HSQC, HMBC, COSY)

→

Interpret structure & confirm

Anthropic targets the interpret step — pairing "Agent Skills" with agentic tooling to retrieve and read spectra by chemical name or SMILES.

Upside

Could sharply shorten routine ¹H/¹³C confirmation, raising hopes for automating the synthesis-verification loop.

Caveats

General LLMs still trail specialist tools on simulation & docking; full NMR automation needs better image accuracy, and the model now follows instructions more literally.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.