Ratio of AECI frontier-model growth rate after October 2025 to before October 2025, across all AI developers
- Measures
- Ratio of the linear time-trend slope of AECI (Anthropic's version of the Epoch Capabilities Index) scores for frontier AI models after a trend break around October 2025 to the slope before it, fitted across frontier models from all developers rather than Anthropic models alone.
- Population
- Frontier AI models from all developers scored on AECI, including Anthropic internal-only models, through October 2026
- Conditions
- Piecewise-linear OLS fit with a single breakpoint near October 2025, as in METR's November 2026 assessment; acceleration reported as post-break slope divided by pre-break slope.
- Unit
- x
Most likely about 1.40 x; 90% interval 1.10 to 2.10 x.
Stated as a distribution, the precision the evidence supports. right-skewed; the frontier-wide fit itself is near 1.4, with wider uncertainty from link-function and composition choices
Assessment
The graph's estimate, all things considered: the figures and arguments it rests on, and how they were combined.
The only published figure for this ratio comes from METR's November 2026 independent assessment of AI R&D acceleration at Anthropic. When METR re-ran its main AECI analysis on all frontier models rather than Anthropic models alone, it estimated about 1.4x acceleration after October 2025, against 1.9x (95% interval 1.5 to 2.8x) for Anthropic alone. METR gives no interval for the 1.4x figure.
The figure carries a compositional caveat that METR states itself: the historical trend is fitted mostly on non-Anthropic models and the recent trend mostly on Anthropic models, several of them internal-only and not publicly benchmarked. So the ratio partly compares two developers' trajectories. METR offers two readings. If Anthropic was behind the frontier in 2025, the frontier-wide 1.4x is the more accurate measure of acceleration. If other developers' earlier models were more tuned to benchmarks, the pre-break frontier slope is overstated and 1.4x understates acceleration.
The estimate is centred on METR's 1.4x. Its 90% interval of roughly 1.1x to 2.1x reflects three things: the missing reported uncertainty; the large sensitivity of the Anthropic-only estimate to the choice of link function (1.1x to 2.9x); and the composition issue. The underlying data are partly confidential, so the figure rests on METR's testimony about evidence not shown. Anthropic had an interest in how the report came out and held redaction rights over it.
The work: which figures, how they were combined, and the arithmetic
Source figure: 1.4x point estimate (METR, 2026-11-05, AECI appendix, robustness section 'AECI shows less acceleration when we include non-Anthropic models'). No interval is reported. As a scale for sampling and fit uncertainty I used the Anthropic-only interval (1.5 to 2.8 around 1.9, about -20%/+45%). Applied to 1.4, that gives roughly 1.12 to 2.0. I widened the upper tail slightly for the link-function sensitivity (cauchit and log-log links push the Anthropic-only figure to 2.8 to 2.9) and for the benchmark-optimization reading METR offers. Result: 5th percentile 1.1, median 1.4, 95th percentile 2.1. A ratio below 1 is implausible, given that frontier ECI trends are publicly observed to have steepened since late 2025.
Estimates
Each figure a source states for this quantity, in its own words, under the argument it belongs to and read against the graph's estimate.
Grey band: the graph's 90% interval, with its central value as a line. Dark marks: each source's own figure. A thin line is an interval, a thick bar its middle half, a dot a single figure, a dashed arrow a bound. An open box is where an argument places the value.
Decomposition
This claim is atomic: it bottoms out in a bedrock fact, an open empirical question, or a value premise, and does not decompose further.
Provenance
Where this quantity's evidence comes from. The figures sources state are under Estimates above.
Its one source has been traced to an origin, or to one another.
If we take our main AECI results and plot all frontier models instead of just Anthropic models, we estimate a lower degree of acceleration.
Appendix: all-developer frontier AECI trend rises from +17 to +24 points per year, a 1.4x acceleration.
Cite this claim: a formal citation with its evidence attached
Embed this claim: show it on another site with its current assessment
An embed reports the graph's judgment as it stands when the page is viewed: the status with its confidence, how disputed the claim is, and the date. It links back here.
<blockquote class="minerval-claim" data-claim-id="f88b4b60-915f-4e78-a20b-59a370cdd0fe"><a href="https://minerval.ai/claims/f88b4b60-915f-4e78-a20b-59a370cdd0fe">Ratio of AECI frontier-model growth rate after October 2025 to before October 2025, across all AI developers</a></blockquote> <script async src="https://minerval.ai/embed.js"></script>
<iframe src="https://minerval.ai/embed/claims/f88b4b60-915f-4e78-a20b-59a370cdd0fe" title="Minerval: Ratio of AECI frontier-model growth rate after October 2025 to before October 2025, acros…" width="480" height="250" style="border:0;max-width:100%" loading="lazy"></iframe>
[](https://minerval.ai/claims/f88b4b60-915f-4e78-a20b-59a370cdd0fe)
The badge now: 
Contribute
Every judgment on this page is open to challenge. A contribution is evaluated on its merits by the reviewer; if it succeeds the page changes, and if it does not, the reasons are stated. Either way the exchange becomes part of the claim’s public record.
Stated views None yet. A view stated to the graph is shown on this page in its speaker’s words, as a source like any other, and carries no weight in the assessment.
Created by extractor · Nov 13, 2026. Every judgment on this page is accompanied by a reasoning trace.