Unverified50% confidenceBenchmarkExact time
Against GPT-oss-120B, GPT-5.4, Claude-Haiku-4.5 and GPT-6-Astra, the adversarially trained SEMA-Auto achieves an average of 59.6% ASR@1 on HarmBench.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/27/2026
First Seen
Valid until: 1/25/2027
Sources
Related Entities
Related Claims
VerifiedThe API model identifier for GPT-6 Astra is gpt-6-astra.70% similarUnverifiedUnder the medium-intensity setting of the Agents' Last Exam benchmark, GPT-6.1 Terra's score is on par with GPT-6 Astra.70% similarVerifiedGPT-6 Astra is an entirely new pre-trained model generation, not a minor version iteration.69% similarUnverifiedIn the Exploit Gym benchmark, the previous-generation GPT-5.6 Sol achieved a completion rate of approximately 30%, while GPT-6 Astra reached a success rate of approximately 40% with a significantly reduced token consumption.69% similarUnverifiedGPT-6 (Astral) achieved an accuracy rate of 72.6% on the OS World 2.0 benchmark, with an average of approximately 40 minutes per task.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1103177API
curl https://kongchang.com/api/v1/knowledge/claims/1103177MCP
get_claim(id=1103177)