Inherent, a San Francisco-based startup founded by former DeepMind researchers, is claiming its AI teammate outperformed leading frontier models from Anthropic and OpenAI on a research replication benchmark.
What Happened
According to the company, Inherent has developed an AI system it describes as an autonomous "research teammate." The company reports that this system achieved higher accuracy than comparable offerings from both Anthropic and OpenAI when tested on tasks involving the replication of published scientific research. The startup was founded by alumni of Google's DeepMind division, a leading artificial intelligence research laboratory.
Why It Matters
Research replication represents a demanding test of AI reasoning capabilities, requiring models to understand methodology, execute complex experimental procedures, and validate results against original findings. If verified, strong performance on this task would suggest meaningful progress toward AI systems capable of contributing meaningfully to scientific discovery. The ability to reliably replicate research could accelerate scientific workflows and help identify irreproducible studies.
The Bottom Line
Inherent's claims have not been independently verified, and the company has not yet publicly released benchmark results or technical details supporting its assertions. As with any vendor capability claim, independent evaluation will be necessary before the performance differential can be confirmed.