AI's two clocks

The fast clock · global

650× cheaper

GPT-4-level intelligence cost $37.50 per million tokens in Mar 2023. Today 130 models from 14 labs match it, 56 of them open-weight, from $0.058. See every level

Capability frontier (ECI): 167.3; Claude Opus 5.5, Anthropic · up from 150.0 a year earlier
Longest task done half the time: 17h; Claude Mythos Preview (provisional, above METR's 16-hour limit) · 1740× longer than at ChatGPT's launch; When could it do yours?
Open-weight lag: 4.4 months; Kimi K3 (Moonshot) matched GPT-5.4
Releases in the last 90 days: 49; 29 independently scored so far

Source: Capability: Epoch Capabilities Index. Prices: providers' list prices, blended 3:1 input to output, per token. Task horizons: METR. · Epoch Capabilities Index (ECI) · Data on AI models · LiteLLM model price map · Task-completion time horizons (Time Horizon 1.1)

The slow clock · by country

OECD average

Output per worker, year on year: +1.4%; 2025 · 0.9% a year over the last decade
Output per worker since ChatGPT: +2.3%; 2022 to 2025, in total
Unemployment: 5%; Apr 2026, up 0.1 pt since Nov 2022 (was 4.9%)
Youth unemployment, ages 15–24: 11.4%; Apr 2026, up 0.8 pt since Nov 2022 (was 10.6%)

Context, not cause: unemployment moves with interest rates and the business cycle, and productivity with everything else. Output per hour comes from the BLS (US, quarterly); output per worker from the World Bank (annual, comparable across countries).

Source: Unemployment: BLS (US), OECD harmonised rates, Eurostat (EU). Productivity: BLS output per hour (US) and World Bank output per worker. · US Bureau of Labor Statistics · OECD Data Explorer · Eurostat · World Development Indicators

Source: Epoch AI, METR, LiteLLM, BLS, OECD, Eurostat, World Bank · data updated Oct 2, 2026howfastis.ai ↗