
How to read this finding
These are reported model specifications and launch-era prices. Benchmark comparisons depend on task, inference budget, and evaluation setup; they are not a universal ranking of intelligence.
o1 used additional computation while answering and reinforcement learning to improve its reasoning process. The report emphasized gains on complex math and science, with higher cost and slower responses than the general-purpose alternative.

These are reported model specifications and launch-era prices. Benchmark comparisons depend on task, inference budget, and evaluation setup; they are not a universal ranking of intelligence.
Evidence you can use
Historical snapshot: October 2024. Dates and populations are specified per row.
| Measure | Reported value | Definition and source |
|---|---|---|
| Llama training corpus | 15T tokens | Training corpus size reported for the Llama 3 family.2024 report, slide 15 (PDF page 16) |
| Llama 3.1 largest model | 405B parameters | The largest model discussed in the Llama 3.1 release.2024 report, slide 15 (PDF page 16) |
| o1-preview output price | $60 per 1M tokens | API list price in the October 2024 report; a historical price, not a current quote.2024 report, slide 13 (PDF page 14) |
These are reported model specifications and launch-era prices. Benchmark comparisons depend on task, inference budget, and evaluation setup; they are not a universal ranking of intelligence.
Historical snapshot published October 10, 2024. This web edition was prepared on October 10, 2026 from the online deck and original launch posts. Findings and forecasts retain their original time frame.
Benaich, Nathan. “What changed with OpenAI o1 in 2024?.” State of AI Report 2024. Historical report snapshot; web edition prepared 2026-10-10.