epoch.ai
PulseAugur coverage of epoch.ai — every cluster mentioning epoch.ai across labs, papers, and developer communities, ranked by signal.
11 day(s) with sentiment data
-
Claude Opus 4.7 builds 16,000-line toolkit autonomously in 14 hours
Epoch AI has developed a benchmark called MirrorCode to test how well AI models can program autonomously. In a recent test, Claude Opus 4.7 successfully built a 16,000-line toolkit within 14 hours, demonstrating signifi…
-
Claude Opus 4.7 leads AI in code reconstruction benchmark
Epoch AI has developed the MirrorCode benchmark to evaluate AI models' ability to reconstruct complete programs without original code. Anthropic's Claude Opus 4.7 demonstrated strong performance, successfully rebuilding…
-
Cohere CEO: Renting AI from foreign providers is a national security risk
Aidan Gomez, CEO of Cohere, has warned that relying on foreign providers for critical AI services poses a significant national security risk. He highlighted the recent U.S. government action restricting access to Anthro…
-
AI Cost Paradox: Cheaper Tokens Drive Higher Company Bills
Despite a dramatic decrease in the cost per token for AI models, many companies are experiencing rising AI expenditures. This paradox stems from the increased usage of AI, with complex agentic workflows now requiring nu…
-
AI vulnerability discovery to favor defense after bumpy 2026-2027 transition
AI models are poised to significantly enhance vulnerability discovery, shifting the landscape from sparse sampling to dense sampling. This transition, while ultimately favoring defense, is expected to cause a bumpy peri…
-
AI cost paradox: Cheaper tokens drive higher enterprise spend · 4 sources tracked
Despite significant drops in per-token costs for AI models, many companies are seeing their AI expenditures rise due to increased usage and more complex applications. While the cost of AI capabilities has fallen dramati…
-
AI energy use per query drops significantly, new studies show
Recent data from 2025 and 2026 indicates that the per-query energy consumption for AI models has decreased significantly, by an order of magnitude, compared to earlier estimates from 2023. Studies from Google, OpenAI, a…
-
Local LLMs now handle 71% of queries; Xiaomi's MiMo model leads charge
Local large language models have significantly improved, now accurately handling 71.3% of real-world queries, a substantial leap from 23.2% last year, according to Stanford research. This advancement is exemplified by X…
-
Memory costs now two-thirds of AI chip component expenses
Memory has become a critical and increasingly expensive component in AI chip design, now accounting for nearly two-thirds of total component costs. This trend is driven by the escalating complexity and data demands of m…
-
AI foundation model valuations near $1T, but history warns infrastructure builders may lose out
Foundation model valuations are approaching $1 trillion, with Anthropic and OpenAI leading this surge. However, historical parallels suggest that infrastructure providers, like early internet backbone builders, may not …
-
Open AI models lag proprietary versions by 8 ECI points
Epoch AI reports that open-weight AI models have consistently lagged behind proprietary models by an average of 8 ECI points since January 2026. This gap is comparable to the difference observed between OpenAI's GPT-5 a…
-
AI chip costs surge as HBM memory share climbs to 63%
A recent study by Epoch AI reveals that High Bandwidth Memory (HBM) has become the most expensive component in AI chips, with its share of the total component cost rising from 52% in Q1 2024 to 63% by Q4 2025. This surg…
-
AI chip memory costs surge to two-thirds of component spend
A recent analysis indicates that memory components, particularly High Bandwidth Memory (HBM), now constitute nearly two-thirds of the total cost for AI chips. This share has significantly increased from 52% to 63% betwe…
-
AI's power demand shifts from training to distributed inference, challenging grid planning
Utilities are grappling with the significant and evolving power demands of AI, initially planning for massive training campuses but now facing the complexities of distributed inference workloads. These inference demands…
-
Open AI Models Lag Frontier Closed Models, Benchmarks Debated
Several leading AI labs have released new open-source models, including DeepSeek V4, Gemma 4, Kimi K2.6, and MiMo 2.5. An assessment by CAISI suggests these open models lag behind frontier closed models, with the gap wi…
-
Student maps global data center policies amid growing public backlash
A new interactive map created by University of Washington student Isabelle Reksopuro aims to shed light on the growing controversy surrounding data centers. The map tracks data center policies globally, providing a tran…
-
Open AI ecosystems offer cost advantages through shared R&D
The majority of compute costs for developing frontier AI models are attributed to research and development rather than the final training phase. China's AI ecosystem, characterized by its open-first approach among leadi…
-
Google DeepMind AI assists mathematicians, tops FrontierMath benchmark
Google DeepMind has released an AI system called "AI Co-Mathematician" designed to collaborate with human mathematicians on complex problems. This system, built on Gemini 3.1 Pro, achieved a new state-of-the-art score o…
-
Anthropic projected to reach world GDP by 2028 amid rapid growth
A projection suggests that Anthropic's revenue could surpass the entire world's GDP by early 2028. This forecast, originating from Epoch AI and SemiAnalysis, indicates an exponential growth trajectory for the AI company…
-
AI experts clarify AGI timeline definitions, favoring median over mean
A LessWrong post explores the ambiguity in how people discuss Artificial General Intelligence (AGI) timelines. The author suggests that when individuals state a timeline, such as "two years," they often don't specify wh…