OpenAI models' task horizon just 1/12 of METR trend — about 14 months behind, says Toby Ord
tobyordoxford · x · 2026-10-22
Oxford researcher Toby Ord points out that OpenAI models, averaged over Jan–Jul, achieved a task horizon (length of task completable with 80% success, per METR's benchmark) only one twelfth as long as the rapid exponential trend implied — about 3.6 doublings or 14 months behind. METR stopped reporting horizons by April 2026, when the 80% horizon had reached 3 hours.
More from Models
- One researcher pretrains a frontier-class model in 3.5 months for under $10M at Mirendil — vjkaruna · 2026-10-22
- Recursive.SI's Eureka Machine builds Flint-40B-A2B from scratch in 4 months, fast local coding — ChengleiSi · 2026-10-22
- One researcher pretrains 100B-500B coding model in 3.5 months on under $10M, agents write 99% of code — RobertTLange · 2026-10-22
- NVIDIA Nemotron demos go live — mervenoyann · 2026-10-22
- vLLM adds day-0 support for Flint: 500 TPS on a single GB200 — _rockt · 2026-10-22
- WIRED: Recursive's Flint shows AI-built AI is getting closer to real recursive self-improvement — cong_ml · 2026-10-22