Independent Assessment Finds Anthropic AI Development Speed Doubled
An independent monitoring group reports that Anthropic is developing artificial intelligence systems around twice as fast as last year, largely aided by AI itself.
Artificial intelligence developer Anthropic has accelerated its research and development pace significantly over the past year, according to a report released by the independent monitoring group Model Evaluation and Threat Research (METR). The Berkeley, California-based nonprofit was given rare access to internal data and employee interviews at Anthropic as the company prepares for a stock market debut.
The findings highlight a growing trend where artificial intelligence is increasingly being utilized to build subsequent, more advanced versions of itself. While independent external metrics on internal lab progress are rare, the METR assessment provides a rare outside look into the velocity of frontier AI development.
Unprecedented External Access and Findings
METR researchers analyzed internal data and conducted staff interviews across an 11-week period from mid-August through late October. The group estimated that Anthropic is improving the capabilities of its models to handle well-specified tasks around twice as fast as it was a year earlier. According to the report, Anthropic appeared to make as much progress on benchmarks in the past year as it previously projected between October 2023 and October 2025.
The assessment noted that AI itself is playing a major role in accelerating its own development, with the Claude chatbot increasingly assisting researchers. METR researchers also ruled out simply an increase in computing power as the primary driver of this growth rate.
Anthropic co-founder and chief architect Sam McCandlish addressed the findings in a post on X, noting that many of METR's conclusions align with his own views. He stated that the company plans to release its own internal analysis of progress speeds in the coming weeks. However, Anthropic applied redactions to the METR report to mask commercially sensitive details and prevent sharing a roadmap that competitors could use to accelerate their own systems.
Safety Concerns and Self-Improvement
The acceleration report intersects with ongoing discussions regarding safety and governance. Earlier, on September 12, Anthropic CEO Dario Amodei published an essay urging the artificial intelligence industry to slow down capability improvements so that safety measures can keep pace. Amodei highlighted recursive self-improvement—the ability of AI to help build future generations of AI—as a core driver of advancement that requires careful handling.
While METR noted that AI is not yet capable of making fully autonomous and highly significant contributions to the core training process, researchers observed notable "signs of life" indicating that threshold could approach soon. The nonprofit acknowledged caveats in its work, suggesting the acceleration might be limited to well-defined tasks or influenced by historical data baselines, but expressed confidence that development speeds have notably risen.
What it means for developers
As leading labs like Anthropic push the boundaries of model capabilities at an increasing pace, developers looking to integrate state-of-the-art technology into their applications can access top AI models cheaply through one API at https://apixoai.online. This streamlined access allows builders to test and deploy advanced systems without managing multiple provider keys.
Source: Anthropic is reportedly rapidly accelerating AI development amid slowdown calls — Scripps News AI. Written by the Apixo team from that report.
One key for Claude, GPT, GLM, DeepSeek and more. Pay per token with crypto.
Get your API key
