2. Linux.do
  3. 从warp里扒出了模型列表和指标值，分享给大家（更新fable 5、grok 4,5、gpt 5.6系列）

从warp里扒出了模型列表和指标值，分享给大家（更新fable 5、grok 4,5、gpt 5.6系列）

奉刀怀邑武登庸 2026-07-10 03:35 1

不知为何老帖不能编辑了，新发一贴大家看看吧


warp官方很歧视grok 4.5和glm 5.2啊，分低得离谱了
































































































































































































































































































































































































































































































































Model Intelligence Speed Cost
claude fable 5 (max) 0.975 0.400 0.950
claude fable 5 (xhigh) 0.950 0.450 0.950
claude opus 4.8 (max) 0.950 0.400 0.800
gpt-5.6-sol (xhigh) 0.950 0.450 0.800
claude fable 5 0.925 0.500 0.950
claude opus 4.7 (max) 0.925 0.400 0.800
claude opus 4.8 0.925 0.450 0.800
claude opus 4.8 (fast mode) 0.925 0.900 1.000
gpt-5.5 (xhigh) 0.925 0.450 0.800
gpt-5.6-sol (high) 0.925 0.500 0.800
gpt-5.6-terra (xhigh) 0.925 0.450 0.550
auto (genius) 0.900 0.600 0.800
claude fable 5 (medium) 0.900 0.600 0.950
claude opus 4.7 0.900 0.450 0.800
claude opus 4.7 (fast mode) 0.900 0.900 1.000
claude opus 4.8 (high) 0.900 0.500 0.800
claude sonnet 5 (max) 0.900 0.600 0.600
gpt-5.5 (high) 0.900 0.500 0.800
gpt-5.6-luna (xhigh) 0.900 0.500 0.350
gpt-5.6-sol (medium) 0.900 0.550 0.800
gpt-5.6-terra (high) 0.900 0.500 0.550
claude fable 5 (low) 0.875 0.650 0.950
claude opus 4.6 (max) 0.875 0.450 0.800
claude opus 4.7 (high) 0.875 0.500 0.800
claude opus 4.8 (medium) 0.875 0.600 0.800
claude sonnet 5 0.875 0.550 0.600
gpt-5.3 codex (xhigh) 0.875 0.350 0.400
gpt-5.4 (xhigh) 0.875 0.400 0.550
gpt-5.5 (medium) 0.875 0.550 0.800
gpt-5.6-luna (high) 0.875 0.550 0.350
gpt-5.6-sol (low) 0.875 0.600 0.800
gpt-5.6-terra (medium) 0.875 0.550 0.550
claude opus 4.6 0.850 0.500 0.800
claude opus 4.8 (low) 0.850 0.650 0.800
claude sonnet 4.6 (max) 0.850 0.600 0.600
gpt-5.3 codex (high) 0.850 0.400 0.400
gpt-5.4 (high) 0.850 0.450 0.550
gpt-5.5 (low) 0.850 0.600 0.800
gpt-5.6-luna (medium) 0.850 0.600 0.350
gpt-5.6-terra (low) 0.850 0.600 0.550
grok 4.5 (high) 0.850 0.500 0.600
claude opus 4.5 (thinking) 0.825 0.500 0.800
gpt-5.6-luna (low) 0.825 0.650 0.350
claude opus 4.5 0.800 0.600 0.800
claude sonnet 4.5 (thinking) 0.800 0.600 0.600
claude sonnet 5 (high) 0.800 0.700 0.600
gemini 3.5 flash 0.800 0.600 0.250
gpt-5.2 (xhigh) 0.800 0.450 0.400
gpt-5.2 codex (xhigh) 0.800 0.500 0.400
gpt-5.3 codex (medium) 0.800 0.500 0.400
gpt-5.4 (medium) 0.800 0.550 0.550
grok 4.5 (medium) 0.800 0.600 0.600
gpt-5.2 codex (high) 0.775 0.550 0.400
gpt-5.3 codex (low) 0.775 0.600 0.400
gpt-5.4 (low) 0.775 0.650 0.550
claude sonnet 4.6 0.750 0.700 0.600
claude sonnet 5 (medium) 0.750 0.750 0.600
deepseek v4 pro 0.750 0.800 0.250
glm 5.2 0.750 0.850 0.200
gpt-5.2 (high) 0.750 0.500 0.400
grok 4.5 (low) 0.750 0.700 0.600
grok build 0.1 0.750 0.900 0.200
gpt-5.2 codex (medium) 0.725 0.650 0.400
kimi k2.7 code 0.720 0.900 0.180
qwen 3.7 plus 0.720 0.850 0.120
auto (cost-efficient) 0.700 0.500 0.300
auto (responsive) 0.700 0.700 0.600
claude sonnet 4.5 0.700 0.700 0.600
claude sonnet 5 (low) 0.700 0.800 0.600
gemini 3.1 pro 0.700 0.500 0.450
gpt-5.2 (medium) 0.700 0.600 0.400
grok 4.3 (high) 0.700 0.500 0.300
kimi k2.6 0.700 0.900 0.180
qwen 3.6 plus 0.700 0.850 0.150
gpt-5.2 codex (low) 0.675 0.700 0.400
gpt-5.2 (low) 0.650 0.700 0.400
grok 4.3 (medium) 0.650 0.600 0.300
minimax 3 0.650 0.900 0.050
auto 0.600 0.900 0.200
auto (open-weights) 0.600 0.800 0.200
claude haiku 4.5 0.600 0.900 0.200
grok 4.3 (low) 0.600 0.700 0.300
minimax 2.7 0.550 0.900 0.050

最新回复 (2)
  • WolfHolo 07-10 03:46
    1楼

    为什么gpt 5.6系列没用max

  • 夏日渐长 07-10 04:03
    2楼

    这指标应该都是巅峰值吧

    要是前几天的GPT5.5跟豆包做一桌吃饭

* 帖子来源Linux.do
附近帖子
返回
飞读
随机推荐