@bardia_heydari بهش میگن u turn attention, قسمت دوم هم بنچمارک داره لازم نیست با حرف ادم ها بری جلو
مثلا opus 4.8 حتی بالای ۵۰۰ک هم هنوز کویرنس داره البته به خودت کانتنت هم ربط داره
Brruuh = $10
Bruuuuuuuhhh = $20
@AnthropicAI sets Claude Opus 4.8 Fast Mode pricing at $10 per million input tokens and $50 per million output tokens.
By the way, I like the UI.
Thank you for amplifying the voice of the Iranian people with such clarity and care. With or without their support, we will take back our homeland from the grip of darkness.
Europe’s position will be remembered as a permanent shame. Rather than standing with Iranians in their struggle for freedom, they helped preserve a regime built on terror, repression, and bloodshed.
Now there is no room for doubt: their words of sympathy and condemnation were empty gestures. When the moment demanded courage, they chose to legitimize the Islamic Republic, honor its executioners, and clasp hands already stained with the blood of our people.
Whether or not Europe stands with us, whether or not your journalists do their jobs, whether or not your politicians demonstrate the courage to act, I will fight for my people and my country.
Thank you for amplifying the voice of the Iranian people with such clarity and care. With or without their support, we will take back our homeland from the grip of darkness.
Europe’s position will be remembered as a permanent shame. Rather than standing with Iranians in their struggle for freedom, they helped preserve a regime built on terror, repression, and bloodshed.
Now there is no room for doubt: their words of sympathy and condemnation were empty gestures. When the moment demanded courage, they chose to legitimize the Islamic Republic, honor its executioners, and clasp hands already stained with the blood of our people.
I believe the math is obvious here. Anthropic probably bought a certain number of GPUs based on their usage forecasts, but right now demand seems higher than they expected. Because they’re competing with OpenAI, they likely don’t want to cut API usage for customers or raise prices. So would it be rational for them to limit subscription users instead?
They already subsidized usage to encourage adoption, and they may have already achieved that goal. What technical solutions do teams usually use to restrict usage in this kind of situation?
It seems possible that a lot of the highest I/O usage comes from relatively easy tasks. In that case, they could reduce effort per response and make users spend more turns from their subscription without officially changing anything. Power users could still choose higher or maximum effort if they wanted.
@SarcasticPyDev Token sparsity might be affecting performance. You can see a similar pattern in Opus 4.7 vs. 4.6. I cannot speak for DeepSeek, but it could be one possible explanation.
این روایت باید تکرار شه چون نوک کوه یخ جنایات هولناک ماههای اخیره:
به قتل صالحه با شلیک از فاصلهی نزدیک در مقابل چشم همسرش بسنده نکردن. برای باز کردن گوشیش (گوشی یک فرد کشتهشده) رفتن بالای سر جسد صالحه و بهش تعرض کردند و تصاویر این تعرض رو به همسرش نشون دادن.
بشنوید کارتل اراذل اسلامی چه بر سر یک دادخواه آورده:
ماموران جمهوری اسلامی حتی پس از قتل صالحه اکبری حرمت او را نگه نداشتند و با تعرض به پیکر بیجانش، تصاویری را برای همسرش اکبر خدایی ارسال کرده بودند.
آنچه بر این مرد گذشته، فقط اندوه از دست دادن همسرش نبود. شکنجهای عمیق و بیپایان بود. ترکیبی از تحقیر، تعرض به عزیزش و فروپاشی روانی که روح انسان را خرد میکند و چیزی از او باقی نمیگذارد.
احمد خدایی به زندگی خود پایان داد. در واقع جمهوری اسلامی او را هم به قتل رساند.
My take on Opus 4.7 is that, for agentic tasks, it may play a role similar to the one Sonnet 4.6 plays for Opus 4.6 as a subagent. Sonnet is less capable than Opus overall, but it still performs well on well-defined tasks with a clear scope. In the same way, Opus 4.7 may be better suited to serving as a subagent for Mythos on clearly scoped agentic work rather than for long, multi-turn chat interactions, and it may not be designed to rely heavily on subagents itself.