Unverified50% confidenceBenchmarkExact time
在完全相同的模型下，仅靠Harness策略优化，ARC-AGI-3得分从13.3%提升到38.3%，同时输出Token节省约六倍
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
8/26/2026
First Seen
Valid until: 11/24/2026
Sources
Related Entities
Related Claims
UnverifiedOpus 5在ARC-AGI 3上的得分是次优模型的三倍75% similarUnverifiedClaude Opus 5 scored 30.2% on ARC-AGI-3, compared to only 1.5% for the previous generation.66% similarUnverifiedThe o3 model achieved 87.5% on the ARC-AGI test in December 2024, and during the same period reached a rating of 2727 on Codeforces (surpassing 99.95% of human competitors).64% similarUnverifiedOn ARC-AGI-3, Sol Max reached close to 8%, while Opus and Gemini were both below 1%.64% similarUnverifiedOpenAI o4 scored 68.8% on the ARC-AGI-2 test, and GPT-5.2 scored 52.9%.61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/802114API
curl https://kongchang.com/api/v1/knowledge/claims/802114MCP
get_claim(id=802114)