@PawelHuryn: We’ve all been hearing that subscriptions are cheaper than APIs, but I’d never seen the exact number. So yesterday, I t…

We’ve all been hearing that subscriptions are cheaper than APIs, but I’d never seen the exact number.

So yesterday, I took several popular plans and burned a slice of their weekly allowance on purpose. I logged every model call and priced those calls at the vendor’s public API list price.

While I couldn’t test all subscription levels:

According to OpenAI, their monthly allowance is now directly proportional to the subscription price:

• Plus ($20/mo) = the reference OpenAI monthly allowance
• Pro 100 ($100/mo) = 5x monthly allowance - tested
• Pro 200 ($200/mo) = 10x monthly allowance (just downgraded from 20x)

For Claude, their Max 20x is subsidized twice as much:

• Pro ($20/mo) = the reference Claude monthly allowance
• Max 5x ($100/mo) = 5x monthly allowance
• Max 20x ($200/mo) = 20x monthly allowance (not 10x) - tested

According to Meta, for Muse Code:

• Everyday Usage ($5/mo) = the reference Muse monthly allowance
• High Usage ($15/mo) = 5x monthly allowance (not 3x) - tested
• Power Usage ($50/mo) = 20x monthly allowance (not 10x)

What does all this mean? (1/4) 🧵

(2/4) 1. If you want the cheapest western subscription, it’s Grok in the SuperGrok subscription.

2. Just after Grok, Muse Code if you are willing to share your data with Meta (Contributor).

3. Just after Muse Code + Contributor, OpenAI:

a) GPT-6.1 Sol (medium) is 13x cheaper than Grok 4.7 (xhigh) while getting similar scores. SuperGrok subscription multiplier is 18.5x bigger than OpenAI’s.

b) Despite Anthropic’s 2-4x bigger multipliers (depending on the plan), GPT-6.1 Sol will still cost you less than Opus 5.5. A linear scale from my tests:

(3/4) How to use Grok Build or Muse Code without the CLI?

Neither Grok Build nor Muse Code has a native GUI. So, I created one. It currently supports Grok Build, Muse Code, Codex, and Claude Code:

• Extension for VS Code (32K installs): https://marketplace.visualstudio.com/items?itemName=PawelHuryn.grok-vscode-phuryn…
• Extension for Cursor, Antigravity (134K downloads): https://open-vsx.org/extension/PawelHuryn/grok-vscode-phuryn…
• Desktop app for Windows, macOS, and Linux: https://afkpilot.com/desktop

(4/4) Unlike OpenAI or Anthropic, small Grok and Muse plans ($30 SuperGrok, $15 Muse) are enough to do real agentic work. SuperGrok $30/mo also gives you access to the Grok Bot.

We will discuss Bots and Dots another time ;)

All the data shared in this thread: https://github.com/phuryn/experiments/tree/main…

Let me know if I calculated anything wrong! 🙏

P.S. More about my recent benchmarks here. No paywall: https://productcompass.pm/p/ai-subscription-api-value…

I’ve dropped the Max 5x row. My mistake. Anthropic promises “up to 20x usage,” but this applies only to the 5-hour limit.

Keeping what’s measured, as Anthropic doesn’t provide more information. Thank you @blueblimpms 🙏

Do you have a different number? What’s your multiplier? Where was the last time you measured it?

I’m not claiming it’s universal. Its what’s I measured on my machine. But it’s not that far from my daily experience.

I use Grok 30$ as a code reviewer (works for hours most days), Claude 200$ as manager, Codex 100$ as a coder. I hit limits with Claude and Codex only.

Oh wow. Interesting. Would they hide that multiplier? $100 should be 5x $20.

I measured how fast my weekly allowances were depleted and compared this against the reported tokens (input, output, cache) at the official API prices.

All subscriptions were isolated for those tests.

It’s starting point is “Dollar/credit pools priced at public rates and metered APIs….”

They don’t have a frontier model I can measure.

I measured $200, $100 is calculated based in those ratios. If that’s the case, then $100 gets the same multiplier.

Let me verify.

Looks correct, though, I don’t know much about OpenCode.

I tested the multiplier for Max 20x (~45.3 compared to API prices). This number was verified.

For other Anthropic multipliers, I’m sorry, I made a mistake: https://x.com/PawelHuryn/status/2105718604994138422?s=20…

Please treat only Max 20x as verified.

That’s the plan. Can’t wait!

Here:

From tokens the original CLIs report based on the public API prices

Yes