The situation where several companies can produce good reasoning models will rapidly cease as everyone moves further up the scaling curve.
Resolved false The record shows it did not happen or was not so.
Prediction: Statements about events outside the person's control, checked once they fell due: did they come true?
The second judge would not resolve this claim; the verdict shown is the first judge's.
What would count as true or false
The claim (October 2022) is that the situation where several companies can build good, competitive reasoning models will 'rapidly' end as scaling continues, leaving only a few able to compete at the frontier. 'Rapidly' is read as within about two years, so the judgment date is end of 2024. True: by end of 2024 only one or two organisations could field frontier-competitive models, and other companies had clearly fallen behind or dropped out. Mostly true: the frontier narrowed to about three organisations, with everyone else clearly behind. Mixed: the field narrowed somewhat (some players dropped out), but several labs stayed broadly competitive. Mostly false: the number of capable model builders stayed roughly the same, with only slight consolidation. False: by end of 2024 more companies, or about as many (including open-weight and non-US labs), produced models competitive with the frontier on standard benchmarks and leaderboards.
Set before the research began.
Evidence (4)
-
E1 Technical Performance Stanford HAI, AI Index Report 2025
“The AI landscape is becoming increasingly competitive, with high-quality models now available from a growing number of developers.”
-
E2 Technical Performance Stanford HAI, AI Index Report 2025
“Likewise, the difference between the top two models shrank from 4.9% in 2023 to just 0.7% in 2024.”
-
E3 Artificial Analysis AI Review - 2024 Highlights Artificial Analysis
“2024 saw multiple labs catch up to OpenAI’s GPT-4, and the emergence of the first”
-
E4 Open-Weight AI Frontier exponentialtracker.com (Epoch AI Capabilities Index data) · 6 July 2026
“Downloadable weights first passed 130 with Microsoft’s Phi-4 in December 2024, under six months after Claude 3.5 Sonnet shipped”
Reasoning
The Stanford AI Index 2025 says high-quality models were available from "a growing number of developers" [E1], and that the difference between the top two models shrank from 4.9% in 2023 to 0.7% in 2024 [E2]. Artificial Analysis reports that in 2024 multiple labs caught up to OpenAI's GPT-4 [E3]. Epoch-based data shows downloadable-weight models first passing a capability score of 130 with Microsoft's Phi-4 in December 2024, under six months after Claude 3.5 Sonnet shipped [E4]. By end of 2024 the evidence shows high-quality models from a growing number of developers and multiple labs catching up to GPT-4 [E1][E3], rather than a narrowing of the field, which points to the false band of the criteria.
Cite this claim
Dario Amodei, “The situation where several companies can produce good reasoning models will rapidly cease as everyone moves further up the scaling curve.” (said 13 October 2022). Verdict: resolved false. Reliability API, claim c_c30aca4a9cf7, method 0.1.0+c876030c, dataset release v1.1.2. https://epistacklabs.com/claims/c_c30aca4a9cf7
This claim is in dataset release v1.1.2, with the same verdict.
Reply to this claim
More prediction claims by Dario Amodei
-
Mostly false
“We believe AI for biology is on a similar exponential trend.”
-
Mostly true
“We're gonna be fine. The-- the impact of this designation is fairly small.”
-
Resolved false
“where AI is writing 90 percent of the code”
Within three to six months of March 2025, AI will be writing 90 percent of code.
-
Resolved false
“And then in twelve months, we may be in a world where AI is writing essentially all of the code”
Within twelve months of March 2025, AI may be writing essentially all of the code.
-
Mostly true
“We may actually hit that this year”
Verdicts are made automatically; how verdicts are made.