Perceptron Mk1 is a proprietary, API-only video and embodied-reasoning model from Perceptron Inc., positioned as a production layer for "physical AI" use cases where understanding events across time matters. Third-party coverage frames the launch as targeting video understanding workflows beyond chat, including robotics, industrial inspection, sports clipping, and security operations, and characterizes the model as a closed release rather than an open-weights drop. The model is reachable through aggregator routing such as OpenRouter under the perceptron/perceptron-mk1 slug, where it is listed alongside other frontier offerings for unified access.

The model's main selling point, as reported in third-party press, is cost efficiency for video analysis: Progressive Robot's coverage cites a VentureBeat report claiming Perceptron Mk1 is roughly 80–90% cheaper than comparable Anthropic, OpenAI, and Google video-analysis offerings, while still being aimed at enterprise operationalization rather than research demos. In practice, this makes Mk1 a fit for teams that need to ingest live or recorded video streams and want a cheaper backbone for downstream tasks like scene understanding, event detection, or action segmentation, especially where the workload is high enough that frontier video-model pricing becomes a blocker. Because the model is closed and reached through a gateway, integration is straightforward for teams already using aggregator-style APIs, but customization, fine-tuning, or local deployment is not part of the offering.