iFANN
    搜索 iFANN...
    登录
    首页
    新闻
    视频
    图片
    GIF
    探索
    投票
    大奖
    iFAMOUS
    维基
    动漫
    聊天室
    通知
    私信
    收藏
    我的
    维基大奖iFAMOUS排行榜行业创作者奖励用户奖励条款隐私社区准则下架 / DMCA帮助开发者

    © 2026 iFANN

    首页
    搜索
    私信
    提醒
    我的
    照片
    Hype Castle27
    Hype Castle27@hype_castle271w
    📱GPT💭AI💭artificial intelligence
    GPT 6 Astra effort comparison table

    @hype_castle27ran the exact same prompt through GPT 6 Astra in Codex across six distinct effort levels and the numbers tell a weird story about where the value actually hides. low effort spit out 7,923 tokens in 4 minutes for $0.63 while medium pushed to 9,085 tokens in 5 minutes at $0.69. high effort jumped to 19,596 tokens over 10 minutes costing $1.21 and extra high hit 34,420 tokens in 19 minutes for $1.95. max effort generated the most with 37,241 tokens taking 20 minutes at $2.09 but then ultra effort dropped back down to 27,465 tokens in just 14 minutes for $1.84. the max level utilized nearly five times the tokens of the low level and yielded the best result even though it cost more than double the cheapest option. complete outputs from all levels are available in the subsequent thread if you want to see exactly what each tier produced 👀 #GPT

    查看原帖

    GPT 6 Astra effort comparison table

    @hype_castle27 的照片· Sep 10, 2026· GPT

    关于这张照片

    The image is a screenshot of a dark-themed interface displaying a table of data. The table compares "Effort" levels, ranging from "Low" to "Ultra," with corresponding "Output Tokens," "Time to completion," and "Est. Cost." The mood is technical and informative, likely related to AI or software performance. A notable detail is the colored bars next to each effort level, visually representing the output token count. The top of the table has the title "Effort comparison" and shows "6 levels Simultaneous 568d5942".

    查看GPT的全部照片阅读GPT维基

    ?

    更多GPT照片

    查看GPT的全部照片
    Gemini 3.6 Flash vs 3.5 Flash same scoreGemini 3.6 Flash vs 3.5 Flash same scoreGPT-5.6 Sol Terra Luna illustrationGPT-5.6 Sol Terra Luna illustrationClaude Max subscription $200/monthClaude Max subscription $200/monthClaude Sonnet 5 pricing vs Opus 4.8Claude Sonnet 5 pricing vs Opus 4.8Trump administration close to restoring Anthropic Fable 5 accessTrump administration close to restoring Anthropic Fable 5 accessClaude for Government uptime 99.93%Claude for Government uptime 99.93%BridgeAgent loops demoBridgeAgent loops demoGLM 5.2 Artificial Analysis Coding IndexGLM 5.2 Artificial Analysis Coding IndexGLM 5.2 ranks #1 on BridgeBench reasoningGLM 5.2 ranks #1 on BridgeBench reasoningGLM Coding Pro plan $64.8GLM Coding Pro plan $64.8Claude Max subscription limitsClaude Max subscription limitsClaude Fable 5 Max Cursor BenchClaude Fable 5 Max Cursor BenchClaude 3.5 guardrailsClaude 3.5 guardrailsClaude Opus 4.7 Frontend DesignArenaClaude Opus 4.7 Frontend DesignArena
    照片
    Hype Castle27
    Hype Castle27@hype_castle271w
    📱GPT💭AI💭artificial intelligence
    GPT 6 Astra effort comparison table

    @hype_castle27ran the exact same prompt through GPT 6 Astra in Codex across six distinct effort levels and the numbers tell a weird story about where the value actually hides. low effort spit out 7,923 tokens in 4 minutes for $0.63 while medium pushed to 9,085 tokens in 5 minutes at $0.69. high effort jumped to 19,596 tokens over 10 minutes costing $1.21 and extra high hit 34,420 tokens in 19 minutes for $1.95. max effort generated the most with 37,241 tokens taking 20 minutes at $2.09 but then ultra effort dropped back down to 27,465 tokens in just 14 minutes for $1.84. the max level utilized nearly five times the tokens of the low level and yielded the best result even though it cost more than double the cheapest option. complete outputs from all levels are available in the subsequent thread if you want to see exactly what each tier produced 👀 #GPT

    查看原帖

    GPT 6 Astra effort comparison table

    @hype_castle27 的照片· Sep 10, 2026· GPT

    关于这张照片

    The image is a screenshot of a dark-themed interface displaying a table of data. The table compares "Effort" levels, ranging from "Low" to "Ultra," with corresponding "Output Tokens," "Time to completion," and "Est. Cost." The mood is technical and informative, likely related to AI or software performance. A notable detail is the colored bars next to each effort level, visually representing the output token count. The top of the table has the title "Effort comparison" and shows "6 levels Simultaneous 568d5942".

    查看GPT的全部照片阅读GPT维基

    ?

    更多GPT照片

    查看GPT的全部照片
    Gemini 3.6 Flash vs 3.5 Flash same scoreGemini 3.6 Flash vs 3.5 Flash same scoreGPT-5.6 Sol Terra Luna illustrationGPT-5.6 Sol Terra Luna illustrationClaude Max subscription $200/monthClaude Max subscription $200/monthClaude Sonnet 5 pricing vs Opus 4.8Claude Sonnet 5 pricing vs Opus 4.8Trump administration close to restoring Anthropic Fable 5 accessTrump administration close to restoring Anthropic Fable 5 accessClaude for Government uptime 99.93%Claude for Government uptime 99.93%BridgeAgent loops demoBridgeAgent loops demoGLM 5.2 Artificial Analysis Coding IndexGLM 5.2 Artificial Analysis Coding IndexGLM 5.2 ranks #1 on BridgeBench reasoningGLM 5.2 ranks #1 on BridgeBench reasoningGLM Coding Pro plan $64.8GLM Coding Pro plan $64.8Claude Max subscription limitsClaude Max subscription limitsClaude Fable 5 Max Cursor BenchClaude Fable 5 Max Cursor BenchClaude 3.5 guardrailsClaude 3.5 guardrailsClaude Opus 4.7 Frontend DesignArenaClaude Opus 4.7 Frontend DesignArena