METR 发布 Claude Opus 5.5 部署前评估摘要

Summary of METR's predeployment evaluation of Claude Opus 5.5

@misc { metr-2026-claude-opus-5-5 , title = {Summary of METR's predeployment evaluation of Claude Opus 5.5} , author = {METR} , howpublished = {\url{https://metr.org/blog/2026-09-22-claude-opus-5-5/}} , year = {2026} , month = {09} , }

Note on independence: This evaluation was conducted under an unpaid agreement for AI R&D assessment. 1 We drafted the initial summary, and then Anthropic had the opportunity to review and edit the text. We signed off on this final text from the Claude Opus 5.5 system card .

Our preliminary evaluation focused on how Claude Opus 5.5 might impact AI R&D, mainly based on its capabilities on difficult, long-horizon tasks. The main claims we attempt to assess in this report are: (A) would AI R&D at Anthropic now be dramatically accelerated by using Claude Opus 5.5; and (B) was AI R&D at Anthropic already dramatically accelerated due to AI during the development of Claude Opus 5.5.

Note that our work was oriented around collecting evidence related to AI R&D capabilities but was not meant to verify claims about compliance with any specific threshold from Anthropic’s policies. This report summary also does not attempt to assess whether Claude Opus 5.5 has or does not have particular alignment properties.

We conducted a preliminary evaluation of Claude Opus 5.5 informed by:

We divide our conclusions into two categories: those that relate to (A) the level of acceleration of AI R&D of which Claude Opus 5.5 is capable, and (B) the level of acceleration of AI R&D due to AI that went into the development of Claude Opus 5.5. Based on the available evidence, we arrive at the following conclusions:

(A) We believe that acceleration from this model would be slightly higher than for Fable 5.1, but that this model is unlikely to be able to fully automate AI R&D. This is because:

We also made use of an additional source of information which we are not able to disclose at this time. We used this source to understand AI R&D at Anthropic, but it did not relate to METR’s understanding of this particular model.

(B) We believe that the development of this model was at least somewhat accelerated by AI but is unlikely to have been dramatically accelerated by AI. This is because:

Our AI R&D assessment agreement with Anthropic includes a transparency provision that lets us disclose certain facts publicly without Anthropic’s consent: the existence and general terms of the review and redaction process, whether Anthropic exercised its redaction rights, whether any of our findings depended significantly on redacted information, the level of access we were given, and the confidentiality framework that applied to the evaluation. ↩

@misc { metr-2026-claude-opus-5-5 , title = {Summary of METR's predeployment evaluation of Claude Opus 5.5} , author = {METR} , howpublished = {\url{https://metr.org/blog/2026-09-22-claude-opus-5-5/}} , year = {2026} , month = {09} , }

document.addEventListener('DOMContentLoaded', function() { document.querySelectorAll('.btn-bibtex').forEach(function(button) { button.addEventListener('click', function() { button.nextElementSibling.classList.toggle('active'); }); }); document.querySelectorAll('.bibtex .copy').forEach(function(button) { button.addEventListener('click', function() { var code = button.closest('.bibtex').querySelector('code'); navigator.clipboard.writeText(code.innerText).then(function() { button.querySelector('i').classList.replace('fa-clipboard', 'fa-clipboard-check');

setTimeout(function() { button.querySelector('i').classList.replace('fa-clipboard-check', 'fa-clipboard'); }, 2000); }); }); }); });

Want to contribute to this work? METR is hiring: View open roles

METR researches, develops and runs cutting-edge tests of AI capabilities, including broad autonomous capabilities and the ability of AI systems to conduct AI R&D.

A survey of 349 technical workers finds a median 1.4–2x self-reported change in value of work due to AI tools, expected to grow over time, though there are reasons to be skeptical of the magnitude.

We show preliminary results on a prototype evaluation that tests monitors' ability to catch AI agents doing side tasks, and AI agents' ability to bypass this monitoring.

We build on our time-horizon work and analyze 9 benchmarks for scientific reasoning, math, robotics, computer use, and self-driving in terms of time-horizon trends; we observe generally similar rates of improvement to the 7-month doubling time in our original time-horizon work.