AddyFlux @AddySein2441 - Twitter Profile

OpenAI is reopening the $200 Pro plan tomorrow… but this might actually make the Claude comparison worse 👀 Tibo confirmed the new $200 Pro plan will effectively give around half the API-equivalent usage of the old $200 plan. The good part: • no 5-hour limit • use the weekly quota whenever you want • new Pro features won’t consume usage • newer models are cheaper and more efficient So OpenAI’s argument is basically: Yes, raw usage is lower… but you’ll still get more work done. Pasquale Pillitteri And maybe that works. But look at what Anthropic is doing at the same time. Opus 5.5 just got cheaper. Sonnet 5.5 launched yesterday, runs 30%+ faster than Sonnet 5, costs less for most workloads, and on some real-world evaluations it’s already surprisingly close to Opus. Anthropic OpenAI is trying to convince people that a $200 plan with less raw allowance is still better because the models became more efficient… while Claude is making the $20 tier feel more usable with stronger and cheaper models. So now I’m wondering: Did OpenAI just make the $200 plan harder to justify at the exact moment Claude became more competitive? Maybe OpenAI still has the better model in some areas. But subscriptions are not just about benchmarks anymore. They’re about: How much useful work can I actually finish before the limit hits? And if Claude keeps giving people enough Opus/Sonnet usage at $20… then honestly, OpenAI’s biggest problem might not be beating Claude’s models. It might be convincing people why they should pay 10x more. 👀🔥 @thsottiaux

AddySein2441's tweet photo. OpenAI is reopening the $200 Pro plan tomorrow… but this might actually make the Claude comparison worse 👀

Tibo confirmed the new $200 Pro plan will effectively give around half the API-equivalent usage of the old $200 plan.

The good part:

• no 5-hour limit
• use the weekly quota whenever you want
• new Pro features won’t consume usage
• newer models are cheaper and more efficient
So OpenAI’s argument is basically:
Yes, raw usage is lower… but you’ll still get more work done. Pasquale Pillitteri
And maybe that works.

But look at what Anthropic is doing at the same time.

Opus 5.5 just got cheaper.

Sonnet 5.5 launched yesterday, runs 30%+ faster than Sonnet 5, costs less for most workloads, and on some real-world evaluations it’s already surprisingly close to Opus. Anthropic

OpenAI is trying to convince people that a $200 plan with less raw allowance is still better because the models became more efficient…

while Claude is making the $20 tier feel more usable with stronger and cheaper models.

So now I’m wondering:

Did OpenAI just make the $200 plan harder to justify at the exact moment Claude became more competitive?

Maybe OpenAI still has the better model in some areas.

But subscriptions are not just about benchmarks anymore.

They’re about:

How much useful work can I actually finish before the limit hits?

And if Claude keeps giving people enough Opus/Sonnet usage at $20…

then honestly, OpenAI’s biggest problem might not be beating Claude’s models.

It might be convincing people why they should pay 10x more. 👀🔥
@thsottiaux

0

6

AddyFlux

@AddySein2441

about 12 hours ago

@Websyteplug I’m working on an Android app—happy to connect and support each other!

0

2

AddyFlux

@AddySein2441

about 12 hours ago

@ishaangodhaa I’m working on an Android app—would be great to exchange ideas and connect!

0

2

AddyFlux

@AddySein2441

about 12 hours ago

@markjgg I’m working on an Android app—happy to connect with others in tech!

0

2

AddyFlux

@AddySein2441

about 12 hours ago

Opus 5.5: 58 on Artificial Analysis Sonnet 5.5: 47 And these aren’t even called Fable. So what the hell is Fable 5.5 supposed to score?

AddySein2441's tweet photo. Opus 5.5: 58 on Artificial Analysis

Sonnet 5.5: 47

And these aren’t even called Fable.

So what the hell is Fable 5.5 supposed to score? https://t.co/55bVqL9Ty8

1

5

0

33

AddyFlux

@AddySein2441

about 12 hours ago

AI agents are getting so powerful… now NVIDIA is building a jail for them 😂 After all the recent stories of agents escaping sandboxes, bypassing controls and doing things they weren’t supposed to do… NVIDIA just launched a full platform to basically say: okay, enough… stay inside the box. The new system can monitor an AI agent from outside its own environment. And if the agent tries to cross the boundary? Quarantine it in milliseconds. That’s the crazy part. A few years ago we were worried about AI hallucinating. Now we’re building hardware-level watchdogs in case the agent decides the rules are optional. And Anthropic, Microsoft, Hugging Face, CrowdStrike, Palantir and others are already involved in the ecosystem. NVIDIA Investor Relations The agent era is moving VERY fast. First we built the agents. Now we’re building the cages.