Anthropic’s Claude Tackles Chemistry with NMR Structure Analysis

Anthropic, the $965 billion private AI giant, has unveiled a significant step forward in applying its Claude AI model to chemistry, with a focus on NMR (Nuclear Magnetic Resonance) spectrum analysis. Detailed in a newly published white paper, the research demonstrates how Claude can match or even outperform established tools like ChemDraw and MestReNova in predicting molecular structures based on NMR data.

NMR spectroscopy is a cornerstone of chemical analysis, allowing chemists to deduce molecular structures from spectral data. However, the process is labor-intensive, requiring manual peak assignments for every compound. Anthropic evaluated three Claude models—Opus 4.7, Opus 4.6, and Sonnet 4.6—against traditional NMR software on 20 compounds, spanning diverse structural challenges. Impressively, Claude’s Opus 4.7 achieved the most accurate hydrogen predictions, with an average error of just ±0.079 ppm, far below the acceptable margin of ±0.20 ppm.

More strikingly, Claude displayed early aptitude in reverse prediction—inferring molecular structures directly from NMR data, a task that existing tools typically leave to human experts. Out of 15 test cases, Opus 4.7 correctly identified all eight simpler structures and achieved a high success rate on denser, more complex molecules when provided additional contextual data like starting materials. This positions Claude as a potential game-changer for small labs and academic chemists who lack access to expensive, specialized software.

Claude’s Edge in Multimodal AI

Unlike traditional systems, Claude benefits from multimodal reasoning, allowing it to process various forms of chemical data—hand-drawn structures, textual descriptions, and spectral outputs—directly. This flexibility could significantly streamline workflows in a field where chemists juggle multiple notations, databases, and software tools. For example, Claude can interpret experimental methods or supporting information in their published formats, saving researchers hours of manual data parsing.

Yet, limitations remain. The dataset for this evaluation was small, with only 20 compounds for forward prediction and 15 for structure elucidation. Additionally, the model has yet to tackle more complex tasks such as handling 2D NMR data or stereochemical configurations, which are critical in advanced synthetic chemistry.

Anthropic’s Strategic Push into Science

This chemistry-focused initiative is part of Anthropic’s broader “AI for Science” program, which aims to expand Claude’s applications into scientific discovery. The move comes at a pivotal time for the company, which recently surpassed OpenAI in valuation after closing a $65 billion funding round in May 2026. Anthropic’s valuation now stands at an eye-popping $965 billion, bolstered by strong enterprise adoption and a $30 billion annual revenue run rate.

These advancements coincide with Anthropic’s confidential filing for an IPO on June 1, 2026, signaling its intent to go public. With its focus on safety and alignment, the company aims to differentiate itself from competitors like OpenAI and Google DeepMind, while positioning Claude as a practical tool for specialized industries such as healthcare, materials science, and now, chemistry.

Looking Ahead

Anthropic plans to refine Claude’s capabilities further, particularly in areas like synthetic route planning, reaction mechanisms, and interpreting chemical literature. The company is also actively seeking collaboration with chemists and researchers to identify additional bottlenecks where Claude could provide value. Interested parties can reach Anthropic directly through its AI for Science program.

While still early in its chemistry applications, Claude’s performance in NMR analysis highlights the growing role of AI in accelerating scientific research. For Anthropic, this is yet another step toward cementing its place as a leader in AI-driven innovation.

Image source: Shutterstock