Epoch AI's research visualized by The New York Times highlights training compute estimates

Summary

The New York Times has utilized data from Epoch AI to illustrate the extensive mathematical computations necessary for training AI models. This visualization draws from the Epoch Capabilities Index (ECI), which compiles a variety of benchmark results, allowing for meaningful comparisons of model capabilities over time. Epoch AI regularly updates public datasets on AI model training compute and benchmarks, providing essential resources for both media and researchers to understand the complexities involved in developing advanced AI systems.

Analysis

Epoch: Epoch AI is an independent research organization dedicated to investigating the trajectory of AI through data collection, benchmarking, and analysis of capabilities, compute, and infrastructure. It maintains large public databases on AI models and recently published reports on automation potential and new benchmarks in October 2026. In this news, Epoch supplies the training compute estimates and research data visualized by The New York Times. GPT-5: GPT-5 is a multimodal large language model from OpenAI released in 2025, with successors advancing the lineup by 2026. It continues to be referenced in capability evaluations and historical context. The news employs it as a calibration point for the ECI scale alongside other models. Claude 3.5 Sonnet: Claude 3.5 Sonnet is an AI model developed by Anthropic, recognized for its capabilities in coding, agents, and knowledge work. It remains a reference point in benchmark discussions and provider pricing in 2026. The news uses it as one of the calibration anchors for the ECI scale. Epoch Capabilities Index: The Epoch Capabilities Index (ECI) is a composite metric developed by Epoch AI that combines scores from dozens of benchmarks into a single general capability scale for comparing AI models over time. It was actively updated with new model evaluations and domain-specific variants as of mid-October 2026. The news centers on explanations and visualizations of the ECI provided by Epoch. Research Data: Epoch AI regularly publishes and updates public datasets on AI model training compute, benchmarks, and capabilities for use by media and researchers. Capability Tracking: The Epoch Capabilities Index aggregates diverse benchmark results to enable consistent comparisons of AI model performance across evolving evaluations. Media Visualization: Outlets such as The New York Times draw on Epoch's research to create accessible explanations of the computational scale involved in developing frontier AI systems.

Categories

aitechmachine_learning

Related sources

View Original Tweet