Benchmarks
Chemistry benchmarks for language models, built to measure whether a model can do the work a synthesis lab depends on.
onepot-Bench 0
Chemistry knowledge and synthesis decisions
Whether a language model can make the decisions needed to execute a synthesis experiment, partly on reactions run in onepot's own wet lab.
- ChemAbacus · molecular properties from a structure
- SynthRefusal · responses to synthesis requests, benign to banned
- SynthBench · reaction outcome and catalyst choice
onepot-Bench 0.5
LC–MS interpretation by language-model agents
Whether an agent can read raw LC–MS data from a reaction screen, on unpublished acquisitions from onepot's own lab, judged against chemist-adjudicated references.
- LCMSFind · did the intended product form
- LCMSElucidate · which other reaction outcomes occurred