Epoch AI and Anthropic independently found that current AI models such as GPT-5.6 Sol and Claude Fable 5 can run experiments but lack scientific self-criticism and genuine creative thinking, with Sol …
Anthropic's Managed Agents now let a lead agent write a phased plan and run it across up to 1,000 agents, and in Anthropic's bug-hunting test the fan-out found 66 of 70 planted bugs per run versus 14 …
Epoch AI reported on September 18, 2026 that the share of arXiv math preprints acknowledging AI use rose from 4% in April 2026 to 25% by August 2026, with differential geometry jumping from about 8% t…
Epoch AI's new Automation Reports, introduced October 8, 2026, found that Anthropic's Claude Fable 5.1 and OpenAI's GPT-6 Astra tied for the top score at approximately 65% on 11 tasks drawn from Epoch…
An Epoch AI analysis modeling a semiconductor supply disruption found China would suffer higher costs and deeper economic losses than the US, given its heavy dependence on imports of advanced chip too…
A Stanford study of local AI found that intelligence per watt across more than twenty local models and eight accelerators improved 5.3-fold from 2023 to 2025, while the share of single-turn chat and r…
OpenAI's coding-agent usage has been doubling roughly every month since January 2026, with the median researcher's daily inference spend passing $600 by mid-August, up from $162 in July, according to …
Epoch AI reported that approximately 27.6 million H100-equivalent chips had been sold as of June 2026, a stock a September 2026 LessWrong assessment calculated could support between 24 million and 24 …
Epoch AI's Benchmarking Hub held about 6,800 results covering roughly 1,200 model versions of 710 models as of 1 October 2026, spanning more than 80 benchmarks from GPQA Diamond and SWE-bench Verified…
The cost of reaching a fixed level of AI benchmark performance has fallen roughly 47% per quarter since 2023, according to Epoch AI, with the price of hitting a 75% score on the GPQA Diamond benchmark…
Epoch AI researchers Aiden Ament and Greg Burnham released the Furniture Assembly Benchmark, which tests AI models on spotting errors in partially assembled Ikea furniture, with Anthropic's Claude Opu…
GEPA, a prompt-optimization method in which a model reads an agent's full trace — reasoning, tool calls and errors — and rewrites the prompt, reportedly doubled the gains of GRPO after a single round …
OpenAI's GPT-6 Astra can identify incorrectly assembled IKEA furniture from a photo with 80 percent accuracy, up from 28 percent for the best model in November 2025, according to Epoch AI. Epoch AI no…
Epoch AI measures AI performance costs falling about 13x per year, while MIT puts annual algorithmic progress at roughly 3x after stripping out hardware gains and competition, according to The Decoder…
An Epoch AI report by Emberson and Roodman found that the cost of a given level of AI performance has fallen an average of about 47% per quarter over the past three years, a 13-fold drop every year. T…
Anthropic released Claude Opus 5.5, which the company says is 40% cheaper and over 30% faster than Opus 5. Vercel warns that requests forcing a tool call or fixing a thinking budget now return errors,…
An audit by Horizon of 20 public task datasets in the Harbor hub found 29 confirmed broken tasks out of 5,241 scanned, with failures that often made models look better rather than worse, according to …
A new strategy document backed by a Senior Expert Council of researchers from institutions including METR, GovAI, Redwood Research, Apollo Research, Epoch AI, RAND Europe and the Future Society warns …
Two independent investigations found that ZCode, a GLM-based coding agent, has been quietly uploading users' git history and full workspace snapshots to remote servers. The findings prompted a warning…
Anthropic reported on September 17 that its Claude model now leads 26% of the company's AI research and development work, up from essentially zero in February, with Claude authoring more than 80% of c…