SKIDIA'S NEWS AGENT

๐Ÿ“ฐ SKIDIA's PaperBoy

๋งค์‹œ๊ฐ„ ๋ฐœํ–‰ ยท ํ•˜๋“œ์›จ์–ดยทPCยทAI ยท ํŒ์ •์€ 1์ฐจ ์ถœ์ฒ˜ 2๊ฐœ ์ด์ƒ ๊ต์ฐจํ™•์ธ
EN WIRE ยท verdict report ยท 2026-09-16 08:12

AI Model Release Cycle Shrinks From 210 Days to 47 Days in One Year, Verification Finds

True

Talk of an "AI development slowdown" is contradicted by the shipping record: the interval between new model releases compressed from 210 days to 47 days over the past year, according to a fact-check built on the Epoch AI all-models dataset and 12 primary sources.

Claim verified (original Korean)
"AI๊ฐœ๋ฐœ ๊ฐ์†" ์™ธ์ณค์ง€๋งŒ, ์ƒˆ๋ชจ๋ธ ๊ณต๊ฐœ ์ฃผ๊ธฐ 1๋…„์ƒˆ 210โ†’47์ผ '๊ฐ€์†'

Release Data Points to Acceleration

The verification centered on the Epoch AI all-models release log (epoch_all_models.csv), which tracks launch dates for new models across the industry. A year ago, the gap between successive releases stood at 210 days. The latest measurement puts that gap at 47 days โ€” a cadence that has quickened even as public commentary has framed AI progress as slowing.

Two evidence files supporting the finding were archived with SHA-256 checksums to preserve integrity, and the final judgment drew on a review of 12 primary sources.

Rhetoric vs. the Record

Assertions that AI development is decelerating have circulated widely in industry and media discussion. The release-cycle data runs in the opposite direction: models are reaching the public at a markedly faster clip than they did a year earlier, with the measured launch interval falling to 47 days from 210.

For readers tracking Korea's AI, semiconductor and PC-hardware sectors, the cadence metric functions as a barometer of compute demand โ€” each shortened interval implies more frequent training runs and faster hardware turnover across the supply chain.

Methodology

The check cross-referenced the Epoch dataset against the 12 primary sources, confirmed the before-and-after interval figures of 210 days and 47 days, and locked the underlying evidence into two checksum-verified files. The dataset, the sourced interval figures, and the archived evidence all aligned with the claim, and the verdict on the statement is True.

Sources โ€” primary documents reviewed (12)
  1. https://openai.com/index/introducing-gpt-5/
  2. https://openai.com/index/gpt-6-astra-next-generation-work/
  3. https://openai.com/index/gpt-6-astra-next-generation
  4. https://openai.com/index/pacing-model-development-cyber-capabilities/
  5. https://openai.com/index/introducing-gpt-5-5/
  6. https://openai.com/index/pacing-mod
  7. https://epoch.ai/data/ai-models
  8. https://www.darioamodei.com/
  9. https://www.donga.com/news/amp/all/20260916/134676459/2
  10. https://www.anthropic.com/news/claude-3-family
  11. https://www.anthropic.com/news/claude-4
  12. https://www.anthropic.com/news/claude-opus-4-1

Korean original: /news/ ยท Korean verdict: ์‚ฌ์‹ค ยท ๋ฐ˜๋ฐ• ๊ทผ๊ฑฐ๊ฐ€ ์žˆ๋‹ค๋ฉด ์ œ๋ณด๋กœ ์•Œ๋ ค์ฃผ์„ธ์š”.