@OpenAI Computer use just got smarter!
الأصلGPT-6.1 Terra: تقديم استخدام سريع وقادر للكمبيوتر بسعر أقل. يفهم GPT-6.1 Terra ما يوجد على الشاشة، ويتخذ إجراءات عبر التطبيقات، ويكاتب بشكل طبيعي، وهو نموذج أكثر كفاءة ليوم عملك بأكمله مع القدرة على إنجاز المزيد منه.
0
0
0
0
التعليقات 21
-
-
@OpenAI When does it reset? Can the 5-hour limit be removed? Let’s crush that Opus 5.5.
الأصل -
@OpenAI OpenAi going down the efficence lane about to get steamrolled by chinese open Models.
الأصل -
@OpenAI Before switching a team to a new model, run 10 of your real weekly tasks through it and your current one. Count the redo rate. If it's under 2 redos out of 10, the cheaper price is worth the move.
الأصل -
@OpenAI Time to put GPT-6.1 Terra to the test! I’ll be trying it with real-world prompts and sharing my results soon. Stay tuned! 🚀
الأصل -
@OpenAI the real benchmark for computer use is not agents last exam. it is one cookie banner, one surprise update popup, and a dropdown that opens upward
الأصل -
@OpenAI @openai are in trouble. Astra makes more mistakes & drifts more than any previous model. Cheaper price just means more mistakes. I have had to cease all projects with chatgpt, the hour by hour drift & destruction is devastating.
الأصل -
@OpenAI 我发现今天的额度消耗的特别快
الأصل -
@OpenAI pleaseeee already spent like $100 on API calls this week
الأصل -
@OpenAI OpenAi could become
الأصل -
@OpenAI you are welcome @thsottiaux https://x.com/claudeoverflow/status/2114965438522183717
الأصل -
@OpenAI
الأصل -
@OpenAI I’d happily wait longer for an answer I don’t have to babysit. A faster model that needs three rounds of corrections is still slower for me. That’s what would get me using Codex again.
الأصل -
@OpenAI The 50% API price cut creates a second clock for builders: how quickly can you re-qualify the cheaper model? A simple example, not an OpenAI cost claim: a $1,000 monthly model bill falls to $500 at unchanged token usage. If migration and testing cost $2,000, the switch takes four months to pay back. Shorter model cycles can make the integration tax matter as much as the token price. That is why I separate model capability from task reliability. Re-run the evals for tool calls, screen actions and recovery, then compare cost per completed task, including retries and human fixes. A cheaper token is useful; a cheaper verified outcome is the product. This is the engineering side of the model-half-life problem I wrote about: https://x.com/OngroundAI/article/2100595730851193061
الأصل -
@OpenAI Literally nobody cares.
الأصل -
@OpenAI computer use 又卷到"更便宜"这一步了，我比较想看的是它在真实桌面上点错按钮的概率，demo 里谁都点得很准 😅
الأصل -
@OpenAI I’ll test your capable compute
الأصل -
@OpenAI The computer use part is what interests me as a non-coder. if it can drive Higgsfield + an editor end to end, my one person pipeline gets a lot shorter. trying it today
الأصل -
@OpenAI OpenAI says GPT-6.1 Terra's standard API prices are 50% lower than GPT-5.6 Terra and GPT-6.1 Sol. Its Oct 26 launch page lists $1 per million input tokens and $5 per million output.
الأصل -
@OpenAI Bilgisayar kullanma yeteneği ucuzladıkça yapay zeka ajanları demo olmaktan çıkıp günlük işin parçası oluyor. Asıl merak ettiğimiz, Türkçe arayüzlerde ve yerel uygulamalarda ne kadar isabetli çalışacağı.
الأصل -
@OpenAI The lower cost is the real story for small businesses. Computer use was impressive but too pricey for daily back-office work; at this level, things like keying supplier bills into old portals start to make sense.
الأصل
스크롤하면 더 불러옵니다