Skip to content
DevMeme
5881 of 7590
AI ML Post #6440 · source on Telegram

The 10x Price Jump to AI Pro Tier

Description

A screenshot of a dark-mode user interface for an OpenAI subscription upgrade page, titled 'Upgrade your plan'. It presents two options side-by-side. On the left, the 'Plus' plan is priced at '$20 USD/month', with a button below it indicating this is the 'Your current plan'. On the right, the 'Pro' plan is priced at a significantly higher '$200 USD/month', with a bright white 'Get Pro' button. The page details the features of each plan, with the Pro plan offering access to more advanced models like 'o1, o1-mini, and GPT-4o' and a special 'o1 pro mode'. The visual joke and point of discussion in the tech community is the steep, tenfold price increase between the consumer-grade and professional tiers. This highlights the high cost of cutting-edge AI model access and OpenAI's strategy to segment the market between casual users and high-demand professional or enterprise clients who require the highest level of performance and are willing to pay a premium for it

Comments

42
Anonymous ★ Top Pick The $20 plan helps you write code. The $200 plan is for when you need an AI that can withstand the existential dread of reading through your legacy codebase without crashing
  1. Anonymous ★ Top Pick

    The $20 plan helps you write code. The $200 plan is for when you need an AI that can withstand the existential dread of reading through your legacy codebase without crashing

  2. Anonymous

    Sure, $200 a month sounds steep - until you remember it’s still cheaper than the Kubernetes cluster you spun up last year to generate haikus for the marketing team

  3. Anonymous

    Remember when we complained about $10/month for GitHub Copilot? Now we're casually dropping $200/month for an AI that still hallucinates package names and confidently explains why your perfectly valid regex is actually summoning Cthulhu

  4. Anonymous

    When your AI assistant costs more per month than your entire AWS bill, you know we've entered the 'compute as a luxury good' era. That $180 delta between Plus and Pro is essentially paying for the privilege of watching o1 think longer before telling you your architecture is still wrong - just with more confidence and token budget

  5. Anonymous

    “Access to o1 pro mode, which uses more compute” is the new -O3 flag - except turning it on opens tickets with Finance, Legal, and Procurement

  6. Anonymous

    Plus vs Pro: pick your bottleneck - tokens per minute or dollars per minute

  7. Anonymous

    $200 Pro plan: Unlimited o1 for when GPT-4o still hallucinates your microservices as a monolith

  8. @GLXBX 1y

    OMFG you really pay for this shit?

    1. @andrei_nik_kolesnikov 1y

      ye, its like ollama with qwq does not exist or something

      1. dev_meme 1y

        Ngl QwQ is top shit

      2. dev_meme 1y

        I still pay for Sonnet though. There’s still nothing is above Sonnet O1/qwq 32b are somewhat close but meeeeh

        1. @GLXBX 1y

          Jessie, what the f are you talking about?

      3. @qtsmolcat 1y

        Yes all of us have the hardware to run ai models locally how'd you know

        1. @andrei_nik_kolesnikov 1y

          I mean, you probably could rent a VPS with a GPU on it, and it might be much cheaper than 200$/mo

  9. dev_meme 1y

    Ahem

    1. @andrei_nik_kolesnikov 1y

      So, o1-preview could start a docker container much better than o1? Am I reading this right?

      1. dev_meme 1y

        yep, this stat make o1 look fucked up Tbh like almost all benchmarks looks like o1-release is messed up shit

      2. dev_meme 1y

        That's why I would be especially curious to try it out since it seems it was trained without any benchmarks in mind

      3. dev_meme 1y

        Or o1-release is just a failure lol

        1. @andrei_nik_kolesnikov 1y

          This may be related https://openai.com/index/openai-o1-system-card/

    2. @farstars 1y

      What does the task "OpenAI API proxy" refer to? And why are there different levels?

  10. @qtsmolcat 1y

    Of vram, no

  11. @farstars 1y

    That leads you nowhere. Even 20 GB doesn't. I haven't yet seen a model that preforms comparable to or above ChatGPT 3.5 with these limitations. Have you?

  12. @farstars 1y

    Idk what's the model size of gpt 4 or o1, but rumors say those take like 100+ GB.

  13. @farstars 1y

    Yes, and that's either 26 GB or 80 GB in size. Not 8 GB.

    1. @andrei_nik_kolesnikov 1y

      https://smcleod.net/2024/12/bringing-k/v-context-quantisation-to-ollama/

      1. @farstars 1y

        Now that sounds very interesting! Gotta test it later on. Thanks for sharing! Glad we had this talk, guys 😁👌

  14. @kekserw 1y

    https://t.me/istorium_net/8458

  15. @ZgGPuo8dZef58K6hxxGVj3Z2 1y

    """""Unlimited""""" within reason. Tf does that mean?

    1. @danosito 1y

      On "plus" subscription you have 50 req/day for mini and 50 req/week for release I think, on "pro" they've deleted these limites

      1. @ZgGPuo8dZef58K6hxxGVj3Z2 1y

        What the fuck can you do with 50 requests?

        1. @danosito 1y

          I use plus subscription, and o1 mini is really good. I didn't find any competitors to solve math and comprog tasks as good as o1 mini/prev/release

        2. @danosito 1y

          Also, i use chatgpt not to do all my work, just some help for me. So, 50 req/day is more than enough

          1. @ZgGPuo8dZef58K6hxxGVj3Z2 1y

            ChatGPT 4 can't even replace numbers in a number array

            1. @danosito 1y

              Chatgpt 4 was like year ago....

              1. @danosito 1y

                4o now is: Free Fast And smart enough for free model

      2. @ZgGPuo8dZef58K6hxxGVj3Z2 1y

        Clearly not because they had to ass the "within reason label". Like my ISP has the "Unlimited" which is actually only 5TB

        1. @danosito 1y

          Maybe it's not unlimited, idk, i use plus, not pro. You can try and pay 200$, only than we will know, is it really unlimited

          1. @ZgGPuo8dZef58K6hxxGVj3Z2 1y

            It says it in the screenshot

            1. @danosito 1y

              And maybe it really is unlimited. You can check if you want

  16. @Stepan_Poznyak 1y

    *for several years

  17. @danosito 1y

    And, don't get me as a corpo defender, but chatgpt o1(mini and release) has 0 competitors in math/compprog. It's literally the best

Use J and K for navigation