X:ARC Prize (@arcprize)· @arcprize·· 5 天前AI 评分
ARC Prize:GPT-6 Sol 在 ARC-AGI-1 得分 92.7%,单任务成本比 GPT-5.6 Sol 低 73%
R to @OpenAI: Similarly, on ARC-AGI-1 with xhigh reasoning, GPT-6 Sol scored 92.7% compared to GPT-5.6 Sol's 97.5%, but was 73% cheaper per task across matched reasoning levels. Full results: https://arcprize.org/results/openai-gpt-6-sol
ARC Prize 公布测试结果,在 ARC-AGI-1 的 xhigh 推理档位下,GPT-6 Sol 得分 92.7%,GPT-5.6 Sol 为 97.5%。在匹配的推理档位上,GPT-6 Sol 每个任务的成本比 GPT-5.6 Sol 低 73%。完整结果见 arcprize.org/results/openai-gpt-6-sol。
来源:X:ARC Prize (@arcprize) · x.lingyaoai.com