Theo-brainUnofficial. Not affiliated with Theo or t3.gg.
Sign inYou
33:33
0:00

Does the $200 Codex plan suck now?

Wednesday, Oct 7 · 34 min · about 9 min to read

Theo-brain watched all 34 min

OpenAI cut the dollar value of the $200 Codex plan roughly in half, and Theo now measures about $2,000 a month, down from up to $12,000. He argues the plan still works if you use only the much cheaper GPT-6.1 Soul, since fewer dollars can still mean more tasks done. For writing code you ship, he still finds the $200 Claude plan with Opus 5.5 impossible to beat.

Briefing is enoughReading these notes should be enough; watch only if you want to follow his pricing reasoning step by step.

Tap any time to jump the video there.

Key takeaways

OpenAI changed how it counts usage on the $200 Codex plan, cutting the dollar value roughly in half. Theo measured about $570 of usage in a week, around $2,000 a month. A few weeks earlier, the same plan got him up to $12,000 a month.

the $200 plan, in effect giving you half as many dollars as you would have had before. That's pretty bad by itself, but I ran the numbers. I only got $570 out of a week of usage, which adds up to around two grand a month. So, is that the end? Did we really just go from 12K a month of usage to 2K and just smile

In Theo's own logs, about 1,800 responses on GPT-6 Astra (the big model) did 766,000 output tokens and cost $534 at API prices, meaning pay-per-use rates. GPT-6.1 Soul did a similar amount of work, about 1,500 responses with similar output, for $38.27. That is more than a 10x drop.

My usage. My GPT-6 Astra usage was for around 1,800 responses. In in these 1,800 responses, it did 766,000 output tokens and it cost $534. It was API prices. 61 Soul didsimilar number of output tokens. $38.27. Do you understand? Roughly the same amount of usage, greater than a 10x decrease. The reason they made these

Theo ran Terminal Bench, a coding benchmark, himself. GPT-6.1 Soul on its max setting got the best score ever. Soul on its 'X high' setting tied Opus 5.5, but cost 96 cents per task versus $13.11 for Opus, about a 13th of the price.

did. Not only was the 61 Soul max run the best score ever, the 61 Soul X high run pretty much exactly tied Opus 5.5, but Opus cost $13.11 per task. 61 Soul cost 96 cents per task. A 13th the price. Do you understand? Your

For writing code you plan to ship, Theo says the $200 Claude plan with Opus 5.5 is impossible to beat. He meant to cancel Claude subscriptions and instead ended up running six Claude accounts.

y'all. If your goal here is to use your subscription to write code that you plan to ship, the insane value you get out of Opus 5.5 on the $200 Claude plan is impossible to beat. I thought I would cancel subs with Claude because it was so much more efficient, but I ended up just pushing my own limits to what I could do, and I ended up with a six Claude account. And if I refresh now,

Ultrafast mode on Astra is extremely expensive. Having it review two pull requests (proposed code changes) cost $600, roughly what an engineer would cost. On the $500 plan, his weekly limit can be gone in 2.1 hours of generation.

I had Astra on ultrafast review two PRs. It cost $600. $600. We're now in the category where it would have been roughly the same cost to pay an engineer. It would have been slower, but I could have paid him. So, Open AI,$500 plan using ultrafast. You get about $1,000 of ultrafast API value per week on that $500 plan, which means from our estimates 2.1 hours of generation. Yeah, my weekly limit can be killed in 2.1 hours. This was a mistake.

One skill here is new to you.

Claude writes, OpenAI's Soul reviews

What it is
Claude does the main coding work. It then spins up GPT-6.1 Soul sub-agents (helper agents) to review that work, with Claude calling Codex through T3 Code.
Why Theo cares
He finds OpenAI models more thorough in analysis, catching deeper issues than Anthropic models do. Most of his Codex usage is either this kind of review or email and computer-use tasks. For those jobs he finds it incredible and gets plenty of value from his $200 subscription.
How to try it
  1. Keep Claude as your main coding agent for building features.
  2. Ask it to hand its finished work to a Codex sub-agent running GPT-6.1 Soul for review.
  3. Compare what the reviewer catches against Claude's own self-check.
if you can do both, I personally get a ton of value from having Claude spin up sole sub-agents to review its work because Open AI models I've still found are a bit more thorough with their analysis and will find things deeper than Anthropic models tend to. So, I do
a wield and the vast majority of my Codex usage is either me using it to like do my email and computer you or it's coming from Claude calling Codex via T3 code in order to review Claude's work. And for those tasks, I have found it to be incredible and I get plenty of

Codex subscription for automatic PR review New

What it is
OpenAI lets you use your Codex subscription almost anywhere, as long as it isn't serving user traffic. Running it on CI, the automated system that checks code changes, to review each pull request is 'mostly allowed'. Charging others for access to your subscription is not.
Why Theo cares
It is a big difference from Claude, whose subscription only works inside Claude Code. What you're allowed to do with a Codex plan goes a lot further.
How to try it
  1. Ask your agent to set up Codex in your CI so it reviews each new pull request.
  2. Keep it for internal work, not for serving your customers' traffic.
want, basically wherever you want, as long as you're not using it to serve user traffic. So if I put my Codex sub in the cloud and then charge you $5 to access my subscription, that's not allowed. But if I use my Codex sub to run on CI to review everyone's PRs when they come in, that is mostly allowed.

Advice

If you stay on the $200 Codex plan, use only GPT-6.1 Soul and never the Max setting. Theo believes that for almost any real-world use, the plan will then feel practically unlimited.

insane cuz you can absolutely, and I do genuinely believe for any even vaguely real-world use case, you're going to get practically unlimited use of a $200 plan as long as you only use 61 Soul and you never touch Max cuz Max makes no sense at all. Do I think this is a good change

If you'd buy the $200 Codex plan mainly to use Astra, the way he once treated Claude as 'the Fable plan', get the Claude plan instead.

does the $200 sub suck now? If you are thinking about it as the Astra plan, the way I used to think of Claude as the Fable plan, yes. If you're buying the $200 plan just to use Astra or you would buy the $200 Claude plan to use Opus and Fable, yeah, go get the Claude plan. But

Hot takes

Both the subscription value and the API price are 'fake'. Neither reflects what running the models actually costs OpenAI or Anthropic.

His reasoning: API prices are set to see what enterprises will pay, with margins historically around 95% over hardware and energy. Burning $100 of tokens costs the lab more like $2 to $5 in compute.

The reality is that not only is this $200 number made up, so is the API price on the right. Both of these numbers are fake, and neither of them represent the actual costs to the companies when they're running these things. The numbereverything else. But the actual compute and energy costs for burning $100 in tokens is closer to like two to five dollars than it is to that $100 number.

The $500 plan was a mistake, not because it exists but because of its timing. It is an 'absolute optical show'.

His reasoning: Telling people the 20x plan is now 10x, but they can buy a 25x plan for 2.5 times the price, angers everyone. These plans are about marketing, not profit, so a plan that hurts the marketing has failed.

The $500 plan was a mistake. Not the existence of a $500 plan. Doing it right now. Absolute optical show. To tell people that the 20x plan is now 10x, but don't worry, you can get the 25x plan if you pay 2.5 times more, pisses everyone off, myself included. This was just athe labs, not from these $200 subs. This is not real profit for them. These are all about marketing. So, if this $500 plan did not help them with marketing, it failed. And I can comfortably say it

Anthropic is 'petty' for locking the Claude subscription to Claude Code.

His reasoning: You can't use it in other agent tools. Even if you try to make Claude Code act like one of them through its instructions, the request fails.

code. And if you think you can be clever and add the instructions to work like a tool like Pi in your Claude code system prompt, good luck. The request will fail. They are that petty at Anthropic

How Theo thought about this

Theo worked out lab margins from public prices. Fable is the same model as Mythos but sells at less than half the Mythos preview price. He reasoned costs didn't drop, so profit per batch fell from about $118 to $43. Roughly half the profit explains why Anthropic gives only half the limit for Fable.

The margin drop is big, but the return drop way bigger. They used to spend 625 in compute and make $118. Now, they spend the same in compute and they make $43. The line I drew here ishere. It's not exact, but $43 is half to a third of $118. They're profiting half as much, so they give you half as much. Makes all the sense in the world, right?

He refused to judge the plan by its dollar value alone. The question he asked was how many tasks the money completes. Soul's lower cost per task means half the subscription dollars can still mean about five times more tasks done.

And the tokens are cheaper. So, cost per task is what we should be looking at and real-world cost per task on something like Terminal Bench 4 is a 13x decrease. So, the dollar value of the sub being half as much ends up still being five times more tasks depending on the work

Sign in to mark what you've watched and keep notes.

Have a question about this one? Ask about this video →

Theo Daily is unofficial and not affiliated with Theo or t3.gg. Written by Claude Opus 5.5 from each video's transcript.