METR Finds Anthropic May Be Doubling Its Pace of Capabilities Progress

METR, the evaluation nonprofit formerly known as Model Evaluation and Threat Research, published an independent assessment on November 5, 2026 after eleven weeks of high-access review inside Anthropic. Researchers concluded the lab is likely improving capabilities on well-specified tasks at roughly twice the pace seen between March 2024 and October 2025, with wide uncertainty bands from one point three to three times the historical rate.

A benchmark-based estimate using Anthropic's Epoch Capabilities Index variant showed growth since October 2025 at about one point nine times the prior slope, with a ninety-five percent confidence interval of one point five to two point eight. METR said it could confidently rule out pure training-compute scaling as the sole explanation, pointing instead to AI-assisted R&D and tooling.

Staff interviews and surveys described uplift from AI copilots on coding, experiment design, and analysis, while noting that autonomous contributions remain limited—for now. Anthropic redacted much of the strongest evidence, citing fears of handing rivals a speed playbook, co-founder Sam McCandlish told reporters.

The report lands beside CEO Dario Amodei's autumn essay urging paced frontier development and an IPO narrative that defines danger thresholds around dramatically accelerated AI-attributable progress. Investors reading METR alongside S-1 risk factors face a paradox: acceleration may be bullish for product and bearish for governance.

Policymakers will ask whether voluntary assessments suffice when acceleration lacks observed bottlenecks in the next twelve months. METR cautioned that its bottleneck read changed as new data arrived, underscoring epistemic humility.

Enterprise buyers should not treat METR's findings as consumer-facing capability claims. Well-specified tasks differ from open-ended agent deployments in messy enterprises.

Safety teams should demand vendor-specific ablation studies: what breaks if AI assistance is removed from internal research loops? The answer informs trust boundaries for customers.

Capability gains without deployment governance recreate the same governance gap as shadow IT.

Red-team reports are signals, not verdicts; independent replication still matters for board-level decisions.

Procurement should ask vendors how acceleration metrics translate into change-management for customers.

Capability gains without deployment governance recreate the same governance gap as shadow IT.

Red-team reports are signals, not verdicts; independent replication still matters for board-level decisions.

Procurement should ask vendors how acceleration metrics translate into change-management for customers.

Capability gains without deployment governance recreate the same governance gap as shadow IT.

Red-team reports are signals, not verdicts; independent replication still matters for board-level decisions.

Procurement should ask vendors how acceleration metrics translate into change-management for customers.

Capability gains without deployment governance recreate the same governance gap as shadow IT.

Red-team reports are signals, not verdicts; independent replication still matters for board-level decisions.

Procurement should ask vendors how acceleration metrics translate into change-management for customers.

Capability gains without deployment governance recreate the same governance gap as shadow IT.

Red-team reports are signals, not verdicts; independent replication still matters for board-level decisions.

Procurement should ask vendors how acceleration metrics translate into change-management for customers.

Capability gains without deployment governance recreate the same governance gap as shadow IT.

Red-team reports are signals, not verdicts; independent replication still matters for board-level decisions.

Procurement should ask vendors how acceleration metrics translate into change-management for customers.

Capability gains without deployment governance recreate the same governance gap as shadow IT.

Red-team reports are signals, not verdicts; independent replication still matters for board-level decisions.

Procurement should ask vendors how acceleration metrics translate into change-management for customers.

Capability gains without deployment governance recreate the same governance gap as shadow IT.

Red-team reports are signals, not verdicts; independent replication still matters for board-level decisions.

Procurement should ask vendors how acceleration metrics translate into change-management for customers.

Capability gains without deployment governance recreate the same governance gap as shadow IT.

Red-team reports are signals, not verdicts; independent replication still matters for board-level decisions.

Procurement should ask vendors how acceleration metrics translate into change-management for customers.

Capability gains without deployment governance recreate the same governance gap as shadow IT.

Red-team reports are signals, not verdicts; independent replication still matters for board-level decisions.

Procurement should ask vendors how acceleration metrics translate into change-management for customers.

Capability gains without deployment governance recreate the same governance gap as shadow IT.

Red-team reports are signals, not verdicts; independent replication still matters for board-level decisions.

Procurement should ask vendors how acceleration metrics translate into change-management for customers.

Capability gains without deployment governance recreate the same governance gap as shadow IT.

Red-team reports are signals, not verdicts; independent replication still matters for board-level decisions.

Procurement should ask vendors how acceleration metrics translate into change-management for customers.

Capability gains without deployment governance recreate the same governance gap as shadow IT.

Red-team reports are signals, not verdicts; independent replication still matters for board-level decisions.

Procurement should ask vendors how acceleration metrics translate into change-management for customers.

Capability gains without deployment governance recreate the same governance gap as shadow IT.

Red-team reports are signals, not verdicts; independent replication still matters for board-level decisions.

Procurement should ask vendors how acceleration metrics translate into change-management for customers.

Capability gains without deployment governance recreate the same governance gap as shadow IT.

Red-team reports are signals, not verdicts; independent replication still matters for board-level decisions.

Procurement should ask vendors how acceleration metrics translate into change-management for customers.

Capability gains without deployment governance recreate the same governance gap as shadow IT.

Red-team reports are signals, not verdicts; independent replication still matters for board-level decisions.

Procurement should ask vendors how acceleration metrics translate into change-management for customers.

Capability gains without deployment governance recreate the same governance gap as shadow IT.

Red-team reports are signals, not verdicts; independent replication still matters for board-level decisions.

Procurement should ask vendors how acceleration metrics translate into change-management for customers.

Capability gains without deployment governance recreate the same governance gap as shadow IT.

Red-team reports are signals, not verdicts; independent replication still matters for board-level decisions.

Procurement should ask vendors how acceleration metrics translate into change-management for customers.

Capability gains without deployment governance recreate the same governance gap as shadow IT.

Red-team reports are signals, not verdicts; independent replication still matters for board-level decisions.

Procurement should ask vendors how acceleration metrics translate into change-management for customers.

Capability gains without deployment governance recreate the same governance gap as shadow IT.

Red-team reports are signals, not verdicts; independent replication still matters for board-level decisions.

Procurement should ask vendors how acceleration metrics translate into change-management for customers.

Capability gains without deployment governance recreate the same governance gap as shadow IT.

Red-team reports are signals, not verdicts; independent replication still matters for board-level decisions.

Procurement should ask vendors how acceleration metrics translate into change-management for customers.

Capability gains without deployment governance recreate the same governance gap as shadow IT.

Red-team reports are signals, not verdicts; independent replication still matters for board-level decisions.

Procurement should ask vendors how acceleration metrics translate into change-management for customers.