frontier lab53 published results
Anthropic
CoBench and operational automation included. Internal tasks are not public; lab disclosure is not independent replication.
Official site ↗Organizations
Public evidence by organization
Published evidence grouped by organization.
frontier lab73 published results
OpenAI
Current AI self-improvement suite plus original MLE-bench and PaperBench. Graph-only scores withheld.
Official site ↗frontier lab11 published results
Google DeepMind
GRB, a qualitative RE-Bench assessment and AlphaEvolve. Different internal metrics cannot rank labs.
Official site ↗independent evaluator4 published results
METR
Independent evaluations, task horizons and productivity studies. Selected historical snapshots, not a live leaderboard.
Official site ↗benchmark creator0 published results
AI4AI-Bench authors
Original benchmark-author report; no independent replication included.
Official site ↗academic group0 published results
Darwin Gödel Machine authors
Historical v1 study; later revisions require separate reconciliation.
Official site ↗frontier lab1 published results
Meta
Only the Llama 3.1 evaluation reported by MLE-bench authors is included.
Official site ↗frontier lab7 published results
Moonshot AI
Only Kimi K3 in the original AI4AI-Bench study is included.
Official site ↗academic group0 published results
Duan et al.
Source-authored research, with classification separated from independent empirical verification.
Official site ↗other0 published results
Weco AI
Source-authored research, with classification separated from independent empirical verification.
Official site ↗frontier lab3 published results
xAI
Organization of Grok 4 in METR's historical TH1 table.
Official site ↗frontier lab1 published results
DeepSeek
Organization of DeepSeek-R1 in PaperBench's original Table 4.
Official site ↗academic group3 published results
Research outcome prediction authors
Original academic preprint authors; no institutional endorsement inferred.
Official site ↗benchmark creator0 published results
PostTrainBench team
Creators of the original full benchmark and its maintained v1.1 leaderboard.
Official site ↗other2 published results
Intology
External Locus submission reviewed for v1.1 compliance by the benchmark team.
Official site ↗frontier lab6 published results
Z.ai
Organization attribution for GLM 5.2; benchmark agent used Claude Code.
Official site ↗academic group0 published results
Cambridge AI Science & Policy
Publishes the September 2026 working paper. Author affiliations do not imply institutional endorsement.
Official site ↗independent evaluator0 published results
Epoch AI
Independent benchmark reviewer; review does not independently replicate every leaderboard result.
Official site ↗other0 published results
Matthew Schwartz / BootLoops
Author-maintained toolkit; the guest article discloses an Anthropic visiting appointment and states BootLoops is not an Anthropic project.
Official site ↗independent evaluator0 published results
P-Zero Research
Source authorship/evaluator attribution; not an evaluated system or a claim of organization-wide performance.
Official site ↗academic group0 published results
AIM authors
Source authorship/evaluator attribution; not an evaluated system or a claim of organization-wide performance.
Official site ↗academic group0 published results
Priced Guidance authors
Source authorship/evaluator attribution; not an evaluated system or a claim of organization-wide performance.
Official site ↗academic group0 published results
Luo, Kasirzadeh and Shah
Source authorship/evaluator attribution; not an evaluated system or a claim of organization-wide performance.
Official site ↗academic group0 published results
EurekaBench authors
Research collaboration credited by this paper; attribution does not imply an independent replication.
Official site ↗academic group0 published results
RRSI authors
Research collaboration credited by this paper; attribution does not imply an independent replication.
Official site ↗academic group0 published results
MASS authors (UC Berkeley and Sakana AI)
Author group for the MASS two-cycle study; affiliation is not independent scientific endorsement.
Official site ↗academic group0 published results
SAEScientist-Bench authors
Authors affiliated with Institute of Automation, CAS and University of Chinese Academy of Sciences; source-reported benchmark evidence.
Official site ↗benchmark creator0 published results
Proximal
Creator of FrontierSWE; this draft indexes the v2 report and a harness comparison, with full benchmark-result records pending separate review.
Official site ↗academic group0 published results
FML-bench authors
Author group for two separate FML-bench preprints; retain each paper and protocol independently.
Official site ↗other0 published results
The Endgame
Publisher of an explanatory essay informed by anonymous frontier-lab conversations. Inclusion attributes a perspective; no official lab endorsement or empirical RSI evidence established.
Official site ↗academic group0 published results
Physical RSI survey authors
Author cohort for the October 21, 2026 physical RSI survey; not a claim of a separate institution.
Official site ↗academic group0 published results
Continual AutoResearch authors
Bibliographic author cohort for the November 3, 2026 preprint, not a separate institution.
Official site ↗