A viral chart predicting open-weights AI models will catch closed frontier models by December 2026 is misleading, as analysis across 18 benchmarks shows the gap varies by task and is not uniformly shr…
Epoch AI and METR released MirrorCode, a benchmark testing AI models on reimplementing entire programs from scratch. Claude Opus 4.7 achieved a 56% score, solving a bioinformatics toolkit in 14 hours …
Epoch AI's new MirrorCode benchmark tests AI models on recreating complete programs without access to original code. Claude Opus 4.7 leads with a 56 percent solve rate, rebuilding a 16,000-line toolki…
METR and Epoch AI released MirrorCode, a benchmark testing AI agents' ability to reimplement entire software programs from source code. Claude Opus 4.6 rebuilt a 16,000-line bioinformatics toolkit in …
Per-token prices for large language models are collapsing, but AI bills are exploding as reasoning models consume far more tokens per task. Uber burned through a year's AI budget in four months, and M…
NVIDIA released Nemotron 3 on December 15, 2025, offering businesses a specialized AI model they can trust. The model series, including Ultra and Super 100B variants, achieved top benchmarks in instru…
Independent evaluation startups fail because talent is better compensated in post-training and application development, the customer base is too narrow, and model optimization pressure renders evals o…
A one-person AI business faces median gross margins of 52% versus 70-80% for mature SaaS, with 6.1% monthly churn on low-priced products and inference costs falling 50x per year. Over 54% of indie AI …
The cost of running a capable AI model fell by roughly 280 times between November 2022 and October 2024, yet the average company's AI bill rose more than 300% over the same period. The paradox arises …
Uber's COO Andrew Macdonald revealed the company exhausted its 2026 AI coding budget in four months, unable to link rising Claude Code token usage to shipped features. The disconnect between token con…
AI-powered vulnerability discovery, as demonstrated by Mythos Preview, is shifting from sparse to dense sampling of software attack surfaces, potentially leaving attackers with fewer zero-day exploits…
Epoch AI has introduced a taxonomy modeled after the U.S. Department of Labor's O*NET system, mapping over 60 discrete tasks in AI research and development to track how quickly the field is automating…
Epoch AI analysis reveals Microsoft, Amazon, Alphabet, Meta, and Oracle are increasing AI infrastructure spending by 70% annually while operating cash flow grows only 23%, potentially causing spending…
The U.S. government shut down global access to Anthropic's most powerful AI models, triggering a global scramble for sovereign AI as countries realize their dependence on U.S. technology is a strategi…
Epoch AI disclosed on May 11, 2026, that an internal audit of its FrontierMath benchmark, a 350-problem test developed with over 60 mathematicians to evaluate AI reasoning, found fatal errors in rough…
Epoch AI projects that compute power for running AI models will surpass compute power for building them by 2030, with nearly half of inference operations shifting to specialized ASICs while training c…