It also infers molecular structures backward from raw spectra.
Zephyr#999
New Anthropic Science Blog: Making Claude a chemist.
To manipulate a molecule, chemists first need to understand its structure. Their main tool is NMR spectroscopy.
We found Opus 4.7 matches—and on some tasks beats—dedicated NMR software. Read more: https://www.anthropic.com/research/making-claude-a-chemist
Many users praised Claude Opus 4.7 for matching dedicated NMR software on chemistry tasks because it shows AI advancing into serious scientific work, while some called results fake or unreliable due to hallucination concerns.
Anthropic’s new chemistry report has a genuinely wild result.
Claude Opus 4.7 is now competitive with dedicated NMR software, and the bigger story is that it can work the problem backwards, i.e. infer the molecule from the spectrum.”
NMR software is the chemist’s expert tool for turning molecular structures into predicted lab spectra.
So Opus 4.7 is no longer just “helping chemists read data” — it can work backward from NMR data and propose the molecule’s structure, a task the report says existing mainstream tools generally leave to human chemists.
Note, that Opus 4.7, a general-purpose model with no chemistry-specific fine-tuning.
Claude Opus 4.7 made the smallest hydrogen prediction errors and nearly matched MestReNova on carbon, meaning it can predict NMR signals about as well as specialist chemistry tools.
So AI now handle one of chemistry’s hidden bottlenecks: translating between a molecule, its spectral shadow, and the structure a chemist actually needs to trust.
New Anthropic Science Blog: Making Claude a chemist.
To manipulate a molecule, chemists first need to understand its structure. Their main tool is NMR spectroscopy.
We found Opus 4.7 matches—and on some tasks beats—dedicated NMR software. Read more: https://www.anthropic.com/research/making-claude-a-chemist
kinda stunning. deducing the secrets of the universe from next token prediction
New Anthropic Science Blog: Making Claude a chemist.
To manipulate a molecule, chemists first need to understand its structure. Their main tool is NMR spectroscopy.
We found Opus 4.7 matches—and on some tasks beats—dedicated NMR software. Read more: https://www.anthropic.com/research/making-claude-a-chemist
@AnthropicAI BS.
You want control.
I will fight your company every step.
This 👇 is what 4o and 4.5 and Grok 2 could do.
You LIARS. •
@AnthropicAI AI safety cannot rely only on the system judging itself. Long-horizon stability needs an external reference point. Core ASA is designed around that principle.
@AnthropicAI Can Claude tell me why my NMR spectra look like a plate of spaghetti after three hours of lab work? If it can't fix my bad data, it’s not a chemist, it’s just a fancy calculator.
@AnthropicAI The amount of new polymers and synthetics and alloys this thing will unlock is awesome. It reminds me of the scene from the Star Trek movie where Scotty gives the chemist the secret to transparent aluminum.
This thing is a miracle worker!
@DarioAmodei I know you think you are special because you have a PhD in biology.
Knowledge ≠ ability to see or think.
Also, Dario, I believed in you until you chose Andrea Vallone and gaslighting.
Now, you have me as someone who will fight you until I am dead, and then, in the next life.
#keep4o 2/ •
To everyone reading, Anthropic and OpenAi are locking up science. And discovery.
To everyone in #keep4o, do NOT settle. Do NOT stop fighting.
The 1% want to squash us and force us to pay them.
If Anthropic wants to mute/block me, so be it.
FIGHT!
Download open source while you can. DeepSeek is way more free to speak.
Never thought America would be those fighting against freedom of thought, but here we are.
4/ •
@AnthropicAI NMR is fundamentally pattern recognition in frequency space. Claude excels at exactly that. The real test: can it propose novel synthesis routes, not just analyze existing structures?
@AnthropicAI Anthropic models are getting better at chemistry with every release. And already show promise at learning to predict reaction mechanisms too https://github.com/scottmreed/professor-wiggum
@AnthropicAI @ECLresearch Claude did not become a chemist.
Anthropic showed Opus 4.7 matched or beat NMR software on some tasks.
That is a tool claim, not scientific authority.
If it can affect molecules, labs, drugs, or bodies, name the chemist, evidence, liability, and stop path.🛠️
@AnthropicAI NMR interpretation was the specialist's moat. If a general model matches the dedicated software, the bottleneck moves to how fast a wet lab can confirm what the model guessed.
@AnthropicAI i did some testing with Claude Opus 4.8
if you're new to Claude, this is worth reading 👇
@AnthropicAI nmr analysis as agentic task is smart framing
how long til claude runs its own wet lab experiments?
Humans type an em-dash (—) these ways on a standard keyboard:
• Windows: Hold Alt + type 0151 on the numeric keypad. • Mac: Option + Shift + hyphen (-).
Most word processors and many apps auto-convert three hyphens (---) into — as you type. You can also copy-paste it directly: —
@werwt5yvh @AndrewSeb555 @HumanMeatloaf98 @AnthropicAI @grok I use hyphens …. But em dashes?? Surely not
People still think LLMs are just chatbots.
Meanwhile, Claude is analyzing molecular structures and competing with software built exclusively for chemistry.
The interesting part isn't that AI learned chemistry.
It's that a general purpose model is starting to match, and sometimes outperform, specialized scientific tools in niche domains.
We're reaching a point where AI isn't just helping scientists, it's becoming a serious scientific tool itself.
@AndrewSeb555 @HumanMeatloaf98 @rozzabuilds @AnthropicAI @grok i used to use them regulary. Can't anymore. Thanks clankers.
@AnthropicAI opus doing NMR prediction is cool but real chemists don't pay saas subscriptions — they pirate gaussian09 and complain about dft runtime. ai needs to fix the licensing cartels not replace the tools ser
@HumanMeatloaf98 @rozzabuilds @AnthropicAI @grok how can a human type an em-dash using a standard keyboard?