TL;DR
METR estimates that Anthropic’s model capabilities on a set of measurable tasks have been improving about 1.9 times as fast as before October 2025.
METR thinks AI-assisted research explains much of the change. That causal judgment is less certain than the measured trend, and much of its supporting evidence was redacted.
Faster research could feed into better research tools. Whether that loop keeps accelerating remains an open question.
Anthropic appears to have made roughly two years of its earlier benchmark progress in one. The researchers who tried to explain why were allowed inside the company to examine internal evidence.
Then Anthropic withheld much of the evidence those researchers considered most informative from the public report.
On November 5, the independent research group METR published an assessment of AI research and development at Anthropic. Its team spent eleven weeks examining internal model evaluations, research data and interviews. METR’s central finding was that Anthropic’s capabilities on well-specified tasks appear to be improving at around twice the earlier rate.
The possibility behind that number is bigger than a faster model release cycle. AI systems are now helping researchers do the work that produces the next generation of AI systems. Better models could make that research faster again.
There are several steps between observing faster progress and establishing that AI is causing a self-reinforcing acceleration. The report is unusually insightful because it gives us evidence for some of those steps and admits how much remains uncertain.
What actually got faster?
METR analyzed scores from Anthropic’s internal models using AECI, Anthropic’s version of the Epoch Capabilities Index. It combines results from several capability benchmarks into one measure. METR estimated that progress was 1.9 times as fast as before, with a reported 95% confidence interval of 1.5x to 2.8x. That interval covers uncertainty in the model scores.
The word acceleration needs a little care here. METR isn’t saying each new model is 1.9 times as capable as the last one. It’s comparing how quickly the benchmark index improved across two periods. If a lab previously needed two years to make a certain amount of measured progress, a doubling of that rate would mean making the same progress in one year.
The report puts the change around October 2025. Its estimate includes internal models that hadn’t been released publicly, which is one reason it differs from the numbers in Anthropic’s published model cards.
The estimate also depends on how you construct the index. METR tried different choices of models, dates and benchmarks, with many results close to 2x. Changing the assumptions about how many benchmark tasks are easy or hard produced estimates from 1.1x to 2.9x. Including models from other leading AI labs gave about 1.4x, which changed the comparison: the earlier trend came mostly from other labs, while the recent trend came mostly from Anthropic.
So there is meaningful evidence that Anthropic’s measured progress sped up. The precise multiplier is less clear. And the index measures a particular collection of tasks, not every kind of reasoning, research or real-world impact.
I’ve written before about evaluating agents you can’t trust yet. A benchmark can tell you something valuable about the tasks it measures without settling what happens when the work is messier. The same caution applies here. Faster progress on AECI is evidence of faster progress on AECI. How far that carries into open-ended scientific work is another question.
A researcher can try more ideas
Imagine a researcher who wants to try a change to a training method. Someone has to implement the experiment, connect it to the existing code, run it, inspect the results and decide whether to keep going. Some ideas take a day or two to set up. Others never get tried because the implementation work isn’t worth the likely payoff.
If an AI coding agent reduces that setup time, the researcher can test ideas that previously would have stayed on the whiteboard. The gain isn’t limited to doing the same work faster. It changes which experiments are affordable to attempt.
In interviews with METR, Anthropic researchers described trying substantially more ideas and shortening the time between an idea and an experiment.
METR also examined training compute, the computing resources used to train models, along with staffing, code contributions, experiment usage and other possible explanations. The team says it found convincing evidence that the acceleration wasn’t caused by faster growth in training compute. It couldn’t publish the underlying details.
After weighing what it saw, METR judged that AI probably accounts for a majority of Anthropic’s current acceleration. It offered a rough guess of about 80% for October 2026, with substantial uncertainty. The report also says non-AI factors likely played an important role in the acceleration that began in late 2025.
The benchmark trend is an estimate from observed scores. The share caused by AI is an informed judgment about what would have happened without AI.
The 12x number is not the 1.9x number
Anthropic staff responding to informal Slack polls estimated an average productivity gain of about 12x from the most capable internal model in October 2026. METR treats the polls as very loose evidence because respondents may not represent the wider staff and the answers reflect little deliberation.
It’s tempting to put that beside 1.9x and declare that the numbers prove each other. They don’t measure the same thing.
The 12x figure comes from employees estimating how much faster they feel with AI. The 1.9x figure estimates the change in a benchmark growth rate. A researcher might get a coding task done much faster without the overall model-development process speeding up by the same amount. Training runs still take time. Experiments can fail. Someone still has to decide which result is worth pursuing.
In a 2025 randomized study, experienced open-source developers believed AI had made them faster. On the tasks studied, allowing AI actually made completion take 19% longer. That result applied to early-2025 tools and those developers’ particular work. METR now says it believes those historical results no longer reflect the current impact of AI models on open-source developer productivity.
METR tried a follow-up with newer tools, then reported in February 2026 that its design was struggling: developers who strongly preferred AI were less willing to participate in work without it, and some avoided submitting tasks they didn’t want to do manually. The researchers judged the newer results too biased to give a reliable overall speedup estimate.
I like that these findings sit beside the Anthropic assessment. The same research organization has evidence of a measured slowdown in one setting and probable acceleration in another. That’s not necessarily a contradiction. The people, tools, tasks and measurements are different.
Independent judgment, evidence we can’t inspect
METR wasn’t simply repeating an Anthropic press release. Its researchers received extensive internal access, examined multiple possible causes and reported disagreements and reversals in their own interpretations. METR says it wasn’t paid by Anthropic.
Anthropic had the right to redact non-public information before publication. METR says Anthropic removed a majority of the evidence it found most informative for its conclusions.
Anthropic said the redactions were “aimed at removing confidential information that could accelerate other AI developers.”
That includes important details about how research productivity was measured, which parts of the training process accelerated and how the team ruled out competing explanations.
I don’t think that makes the report worthless. Access to internal data can produce a better assessment than outsiders could make from public model releases alone. \
It does change what the word independent can reasonably promise. The evaluators made their own judgments, but the public can’t fully check those judgments against the underlying research.
Does faster research keep making itself faster?
Suppose an AI model helps researchers test twice as many ideas. Some of those experiments improve the next model. That model then helps researchers implement experiments that the previous one couldn’t handle.
There is a feedback loop in that description. Whether it keeps getting stronger depends on what else the research process needs.
Better coding assistance doesn’t by itself give the lab more computing power for training. It doesn’t guarantee that a promising result will work in other settings or that a lab can collect the data it needs. A researcher can generate more candidate ideas than the lab has capacity to test. At some point, the limiting step may move from writing the experiment to running it or making sense of the result.
METR looked for those constraints. It didn’t identify a clear bottleneck likely to stop further progress over the next twelve months, but the researchers say their assessment changed repeatedly as they received more information. They don’t claim to have proved that no bottleneck exists.
And METR says AI’s biggest contribution today is still assisting human researchers. Fully autonomous systems aren’t yet making major contributions to the most important parts of Anthropic’s training process, according to the information the team received.
That’s where I would draw the line on the more dramatic predictions. We have evidence that measured model progress has accelerated and a credible, though uncertain, case that AI-assisted R&D is responsible for much of it. We don’t have public evidence that the whole research process can now improve itself without people, or that its rate of improvement will keep rising.
What I’d want to see next is fairly concrete: comparable capability measurements over more model generations, direct records of how much research work AI actually saves and independent access to the evidence connecting the two. I’d also want to know where extra experiments stop producing extra progress.
Resources
METR, “Independent Assessment of AI R&D Acceleration at Anthropic in October 2026”, November 5, 2026. Primary evidence, methods, redactions and limitations.
METR, “Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity”, July 10, 2025. Randomized developer-productivity study.
METR, “We are Changing our Developer Productivity Experiment Design”, February 24, 2026. Follow-up measurement limitations.


