Gamow Labs' new benchmark shows top AI models can't keep a wet lab running
LabRobotBench v0.1: Benchmarking wet lab robotics
LabRobotBench v0.1 evaluates how well frontier AI models troubleshoot and complete wet-lab experiments on simulated robots that mirror real command interfaces. On a held-out subset, GPT 6.1 Sol scored 46.7%, Opus 5.5 scored 23.0%, and Gemini 4 Argon 20.9%. The tasks expose failures like shelving an unfinished assay, blowing cold-storage deadlines, and reporting a measurement that finished 42.5 seconds late. Gamow Labs is building agents to raise experimental throughput.
Watch Opus put the yellow-labeled tray back on the shelf. Its cells have been washed, but the agent never adds the chemical needed for the test or measures the result. The tray is put away; the experiment is left unfinished.