Claude Max costs $100 a month and comes with $100 a month in API credits. Claude Ultra costs $200 and comes with $200 in credits. If you already pay Anthropic for API usage, its top plans hand back their own price.
Pay for ChatGPT Plus at $20, Claude Pro at $20, Google AI Pro at $19.99 and Perplexity Pro at $17 a month billed annually, and you spend $76.99 a month on 4 assistants that cover much of the same ground.
On November 1, 2026 we checked 32 AI plans and 16 API models on each provider’s official pricing page. Then we built a calculator that takes your weekly usage and returns the cheapest stack that still covers it, the same kind of working Excel model as our AI CEO Dashboard Kit.
Premium members get the AI Plan Calculator: enter your weekly usage and see your stack and your yearly savings in 5 minutes.
Here is what the prices say:
The $20 tier carries most of the value for most founders.
The $100 and $200 plans pay off when you run Claude Code every day or build on the API.
A plan beats the API as long as your work fits inside its limits.
Caching and model routing cut token prices by 95% or more.
I ranked 18 of these plans by value for a founder. Claude Pro sits alone at the top because $20 a month includes Claude Code.
1. The $20 tier is where the value sits
Claude Pro includes Claude Code and Research for $20 a month, or $200 a year ($16.67 a month) on annual billing, per Anthropic’s pricing page. Our Claude playbook shows where that $20 pays off first.
ChatGPT Plus also costs $20 and includes Codex and deep research, billed monthly only, per OpenAI. Our GPT-6 Astra masterclass shows how to get more from each prompt.
Google AI Pro at $19.99 puts Gemini inside Gmail and Docs and adds 5 TB of storage, per Google’s plan page. If you already pay Google $9.99 for 2 TB of storage, the upgrade costs $10 more and brings Gemini 4 Argon with it.
xAI now sells Grok and Cursor together with X Premium in one subscription, from $8 to $200 a month. According to xAI, its $30 Premium tier carries the benefits of plans that cost $90 a month combined. Our one-person company manual shows how 3 SpaceXAI engineers started a company in 9 hours.
The free tiers are worth a look too. Kimi gives free users limited K3.1 Pro on the web, and our Kimi playbook covers setup and cost. The DeepSeek app is free.
The rule: pay for one $20 plan in the ecosystem you use most, and add a second plan only for a job the first one can’t do.
2. The $100 and $200 plans pay you back in credits
Claude Max gives you 5x Pro usage every 5 hours and every week. Ultra drops the 5-hour limit for verified accounts and gives 10x Pro usage every week. Both include monthly API credits equal to their price.
Google AI Ultra includes credits too, as Google Cloud credits: $40 a month on the $99.99 plan and $100 on the $199.99 plan.
Here is the math for a founder who spends $100 a month on the Claude API, say to run an agentic OS on Claude. Moving to Claude Max costs $100, the $100 in credits covers the API bill, and the total stays at $100 with 5x Pro usage on top.
ChatGPT Pro starts at $100 and tops out at $500 a month, with a $200 tier in between and monthly billing only, per OpenAI’s help center. OpenAI lists monthly API credits on Pro without publishing the amount.
3. A plan beats the API while your work fits inside it
API prices are public. Claude Sonnet 5.5 costs $2 per million input tokens and $10 per million output tokens, so $20 buys 10M input tokens or 2M output tokens.
Say 4 of every 5 tokens you send are input. A blended million then costs $3.60, and $20 covers about 5.6M tokens a month at Sonnet 5.5 prices.
No provider publishes how many tokens its plans include. Claude says your limits depend on the model you pick and on how long your chats and files are.
So price your real tokens on the API only when you hit your plan’s limit every week, because below that the plan wins. Claude’s paid plans also let you keep working past a limit with usage credits billed at standard API rates.
4. Two settings cut token prices by 95% or more
The first is caching. On Claude Opus 5.5, a cached input token costs $0.20 per million against $4 for a fresh one, a 95% cut. OpenAI’s GPT-6.1 Sol shows the same gap at $0.10 against $2.
The second is routing. Claude Haiku 5.5 costs $0.10 per million input tokens, 40x less than Opus 5.5 at $4. Send classification and formatting to the small model and keep the large one for hard reasoning. Trimming what the agent sends pays too: NVIDIA cut agent API costs by about 50% by redesigning how its agent talks to the model.
The premium edition adds the other token savers with the math, including the price increases already on the calendar.
Keep reading:
You Are Overpaying for Intelligence. Grok 4.5 Just Proved It
Anthropic Just Cut the Price of Frontier Intelligence in Half
What 3 founders save
We ran 3 example founders through the calculator at US list prices, and the price sheet inside it links every source.
The solo founder coder pays $507 a month for 5 plans plus $150 of Claude API usage. The calculator moves him to Claude Ultra, whose credits absorb the API bill, and he saves $3,484 a year.
The non-technical CEO pays $107 a month for 5 plans that overlap. One Claude Pro plan on annual billing plus storage and X Plus brings her to $33, a saving of $884 a year.
The 10-person startup pays $1,341 a month across 4 vendors and the API. Claude Team seats on annual billing bring it to $927, a saving of $4,972 a year.
Inside the premium edition:
The AI Plan Calculator for Excel and Google Sheets: enter your weekly usage and get your recommended stack with your monthly and annual savings
The plan to pay for in 7 founder profiles, plus the 3 example founders broken down plan by plan with what to cancel first
The API vs plan breakeven test with token prices for 16 models, and 10 token savers with the math behind each one
The verified price sheet for 32 plans, with every source link
Get 50% off for 1 year. Premium also opens the full library: the Resources hub, every Claude & Anthropic playbook, the AI Agents builds and the Business & Investing deep dives.