Unverified50% confidenceEventExact time
A specialized model trained with wet lab data outperformed the general-purpose large model GPT-6 Astra on a cutting-edge scientific task.
1
Sources
50%
Confidence
Long-term
Relevance
9/18/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedOpenAI explained that the GPT-6 Astra model itself was not downgraded, and that the degradation stemmed from legacy skill files being triggered too frequently, context management experiments, and inference engine misconfiguration.72% similarUnverifiedAccording to OpenAI's Noam Brown, the model involved is similar in scale to GPT-5.6-level, is a model used solely for internal research and trained to be highly persistent and diligent, and is not the next-generation Astra series.71% similarUnverifiedA new search agent has demonstrated performance surpassing GPT-6 Astra across multiple benchmarks.65% similarUnverifiedTesting of GPT-6 Astra covered various scenarios ranging from code writing, debugging, and refactoring to end-to-end task execution.65% similarUnverifiedThe GPT-5.6 flagship model SO autonomously completed the post-training of the smallest model LUNA end-to-end.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/931717API
curl https://kongchang.com/api/v1/knowledge/claims/931717MCP
get_claim(id=931717)