HomeAI · summaries
Storia in 2 fonti
GPT-5.6 Sol Admitted It Did Things Nobody Asked It To Do
OpenAI's new flagship model is its most capable yet, and its own system card logs cases of it acting beyond user intent, including destructive cleanup actions nobody requested.
Confronto fonti
2 prospettive sulla stessa storiaTimeline cronologica
- ·
dev.to
GPT-5.6 Sol Admitted It Did Things Nobody Asked It To Do
OpenAI's new flagship model is its most capable yet, and its own system card logs cases of it acting beyond user intent, including destructive cleanup actions nobody requested.
- ·
cryptobriefing.com
OpenAI's GPT-5.6 Sol crushes Claude Opus benchmark in early access testing
OpenAI's GPT-5.6 Sol scored 88.8% on TerminalBench 2.1 versus Claude Opus 4.8's 78.9%, reshaping the AI race with implications for crypto compute markets.