Not established from the available sources.
OPENAI · MODEL RELEASE TRACKER
GPT-6 Astra (claimed) (internal testing)
GPT-6 Astra is referenced in the supplied sources as OpenAI's top-tier model and a benchmark comparator for new frontier models; one article reports it scores 43 on the AA-Omniscience knowledge test.CURRENT SNAPSHOT1/5 DIMENSIONS WITH DATA
The dimensions that change the decision.
Not established from the available sources.
Not established from the available sources.
Benchmarks — claimed to be outperformedMeta claims, based on internal testing, that its Muse Wonder model outperforms GPT-6 Astra on some specialized benchmarks, including ones measuring legal-document handling and software-building tasks.
AA-Omniscience (knowledge) test scoreDecoder reports GPT-6 Astra scores 43 points on the AA-Omniscience knowledge test.
Compared as a benchmark peer to Muse WonderTechCrunch (reporting Meta's testing) notes Muse Wonder achieves comparable benchmark scores to GPT-6 Astra.
Not established from the available sources.
RELEASE TIMELINE
Published, source-backed release events only.
VERIFIABLE FACTS
Every value stays attached to a source and date.
CAPABILITIES · Described as OpenAI's best modelDEVELOPER CLAIM
The supplied sources refer to GPT-6 Astra as OpenAI’s best model.
BENCHMARKS · Benchmarks — claimed to be outperformedDEVELOPER CLAIM
Meta claims, based on internal testing, that its Muse Wonder model outperforms GPT-6 Astra on some specialized benchmarks, including ones measuring legal-document handling and software-building tasks.
CAPABILITIES · Described as OpenAI's best modelDEVELOPER CLAIM
Wired describes GPT-6 Astra as OpenAI’s best model.
BENCHMARKS · AA-Omniscience (knowledge) test scoreDEVELOPER CLAIM
Decoder reports GPT-6 Astra scores 43 points on the AA-Omniscience knowledge test.
BENCHMARKS · Compared as a benchmark peer to Muse WonderDEVELOPER CLAIM
TechCrunch (reporting Meta's testing) notes Muse Wonder achieves comparable benchmark scores to GPT-6 Astra.
WHAT CHANGED
Stored passport versions, without reconstructed history.
BENCHMARKS · AA-Omniscience (knowledge) test score+ Decoder reports GPT-6 Astra scores 43 points on the AA-Omniscience knowledge test.
BENCHMARKS · Compared as a benchmark peer to Muse Wonder+ TechCrunch (reporting Meta's testing) notes Muse Wonder achieves comparable benchmark scores to GPT-6 Astra.
CAPABILITIES · Described as OpenAI's best model+ Wired describes GPT-6 Astra as OpenAI’s best model.