True Index: Domain Rankings

Private Top-Tier (S & A) standing across 3 core domains
Agentic & Enterprise72.7p
#12
5 Top-Tier private tests
Coding & SWE72.8p
#11
5 Top-Tier private tests
Math & Scientific Reasoning91.6p
#4
3 Top-Tier private tests

Head-to-Head: Clean Evals vs. Legacy Benchmarks

4 matched domains · positive boost means higher legacy scores
Capability Domain Clean Evals (Tiers S & A) Legacy Benchmarks (Tiers C & D) Public Lift (Boost)
Agentic & Enterprise72.7p (5)63.8p (2)-8.9p
Coding & SWE72.8p (5)76.6p (2)+3.7p
General Knowledge & QA94.7p (3)82.8p (1)-11.9p
Math & Scientific Reasoning91.6p (3)86.2p (3)-5.4p
Non-comparable pre-release tests omitted from audited scoring (model released 2026-09-03): SWE-bench Pro V2 (2026-09-22).
Clean Evals Avg (Tiers S & A)
83.0p
16 private, leak-resistant tests
Legacy Benchmarks Avg (Tiers C & D)
77.3p
8 public or aging tests
Matched Public Lift (Boost)
-0.4p
Near zero or negative indicates clean generalization
Machine-Readable Export

Audited Benchmark Performance of GPT-6 Astra

Grouped by benchmark integrity so private capability signal stands apart from aging public tests.

Top-Tier Validated Tests (Tiers S & A) (16)

Private, leak-resistant tests that build real capability signal.

BenchmarkBenchmark Integrity StatusDomainRankPercentile
MLCR-AAPrivate / FreshMath & Scientific Reasoning#2/4697.6p
FrontierMathPrivate / FreshMath & Scientific Reasoning#4/2793.1p
FrontierCode-AA 2.0Private / FreshCoding & SWE#6/4691.0p
Code2Video-AAPrivate / FreshCoding & SWE#3/2381.2p
MKA (Multimodal Knowledge Agent)Private / FreshAgentic & Enterprise#8/2379.2p
Harvey LAB-AA v1.1Private / FreshAgentic & Enterprise#8/2178.8p
AutomationBench-AAPrivate / FreshAgentic & Enterprise#15/5071.4p
AA-Briefcase v1.1Private / FreshAgentic & Enterprise#17/5067.3p
ARC-AGI-2Private / FreshGeneral Knowledge & QA#2/22100.0p
LiveBenchPrivate / FreshGeneral Knowledge & QA#3/28100.0p
Code Migration v1.1Private / FreshCoding & SWE#4/2484.5p
AA-Omniscience v1.1Private / FreshGeneral Knowledge & QA#9/4784.1p
CritPt v2Private / FreshMath & Scientific Reasoning#3/4084.0p
AA-3DPrivate / FreshCoding & SWE#6/2283.3p
AA-AnalystAgentPrivate / FreshAgentic & Enterprise#9/2467.0p
CyberBench v1.1Private / FreshCoding & SWE#23/2524.1p

Mid-Tier Evaluations (Tier B) (2)

Partially exposed or transitional evaluations with moderate leakage risk.

BenchmarkBenchmark Integrity StatusDomainRankPercentile
HLE-DiamondSemi-ExposedGeneral Knowledge & QA#7/4987.5p
VIALS-AASemi-ExposedMath & Scientific Reasoning#4/2384.1p

Legacy & Vulnerable Benchmarks (Tiers C & D) (8)

Public or aging benchmarks prone to overfitting and training-data leaks.

BenchmarkBenchmark Integrity StatusDomainRankPercentile
GDP.pdfPublic / AgingAgentic & Enterprise#3/5094.9p
Terminal-Bench-Science 0.1Public / AgingMath & Scientific Reasoning#6/4192.4p
BioMysteryBenchPublic / AgingMath & Scientific Reasoning#5/2386.6p
Terminal-Bench 4.0Public / AgingCoding & SWE#9/5083.7p
HealthBenchPublic / AgingMath & Scientific Reasoning#6/1679.6p
SWE Atlas - Codebase QnAPublic / AgingCoding & SWE#7/1569.4p
GDPval-AA v2.1Public / AgingAgentic & Enterprise#28/4332.7p
MMMU-ProPublic / AgingGeneral Knowledge & QA#3/2282.8p