主 题：这样看最夯最实惠的模型似乎是GPT-5.6-SOL--%26gt%3bxhigh
发 布 者：Pokemon
标签分类： 情报
时 间：2026-07-10 20:53:12
内容预览：DeepSWE All Effort LevelsData version: v1.1Tasks: 113Updated: July 9, 2026Source: DeepSWE LeaderboardModelPass@1Avg costOut tokStepsgpt-5.6-sol [max]73% ±3%$8.3960k61gpt-5.6-sol [xhigh]71% ±1%$4.7041k44claude-fable-5 [xhigh]70% ±3%$13.4180k68claude-fable-5 [max]70% ±4%$21.63119k88gpt-5.6-terra [max]70% ±3%$4.9572k76gpt-5.6-sol [high]69% ±1%$3.4728k37claude-fable-5 [high]69% ±1%$9.1857k59gpt-5.6-luna [max]67% ±4%$3.0373k102gpt-5.5 [xhigh]67% ±6%$7.2346k82claude-fable-5 [medium]65% ±4%$6.0940k48gpt-5.5 [high]64% ±3%$5.1031k62gpt-5.6-sol [medium]61% ±2%$1.8618k31gpt-5.6-terra [xhigh]60% ±2%$2.1340k43claude-fable-5 [low]60% ±3%$3.7625k38claude-opus-4.8 [max]59% ±2%$13.22135k120gpt-5.6-luna [xhigh]57% ±2%$1.5445k71claude-opus-4.8 [xhigh]54% ±4%$8.0186k95gpt-5.5 [medium]54% ±3%$2.7520k46claude-s
直达链接： https://www.nodeseek.com/post-816210-1
发 布 者：Pokemon
标签分类： 情报
时 间：2026-07-10 20:53:12
内容预览：DeepSWE All Effort LevelsData version: v1.1Tasks: 113Updated: July 9, 2026Source: DeepSWE LeaderboardModelPass@1Avg costOut tokStepsgpt-5.6-sol [max]73% ±3%$8.3960k61gpt-5.6-sol [xhigh]71% ±1%$4.7041k44claude-fable-5 [xhigh]70% ±3%$13.4180k68claude-fable-5 [max]70% ±4%$21.63119k88gpt-5.6-terra [max]70% ±3%$4.9572k76gpt-5.6-sol [high]69% ±1%$3.4728k37claude-fable-5 [high]69% ±1%$9.1857k59gpt-5.6-luna [max]67% ±4%$3.0373k102gpt-5.5 [xhigh]67% ±6%$7.2346k82claude-fable-5 [medium]65% ±4%$6.0940k48gpt-5.5 [high]64% ±3%$5.1031k62gpt-5.6-sol [medium]61% ±2%$1.8618k31gpt-5.6-terra [xhigh]60% ±2%$2.1340k43claude-fable-5 [low]60% ±3%$3.7625k38claude-opus-4.8 [max]59% ±2%$13.22135k120gpt-5.6-luna [xhigh]57% ±2%$1.5445k71claude-opus-4.8 [xhigh]54% ±4%$8.0186k95gpt-5.5 [medium]54% ±3%$2.7520k46claude-s
直达链接： https://www.nodeseek.com/post-816210-1
