Lately, there have been a lot of disturbing murmurs about the price of AI increasing significantly. As competition increases and power grids tighten, developers are spending more money training and running AI systems. Meanwhile, customers are paying more to get the latest models.
Earlier this week, Anthropic was released in what is expected to be a historic IPO fable 5a dialed-down version of the secretive and supposedly extremely powerful Mythos. Fable 5 is twice as expensive as its predecessor, Opus 4.8, but some users are unhappy with the former’s price. dangerous safety guardrail It becomes effectively unusable in some contexts. Perhaps listening to these fears, OpenAI is now considering significantly lowering the price it charges for its tokens (the basic unit for measuring AI usage). wall street journal reported on Thursday.
For those not familiar with the intricacies of AI finance, this can all be a bit puzzling. It would be very useful if there was an easy way to convert, say, a million “input tokens” into a specific task, but unfortunately there isn’t. Each task has its own computational demands, and a pay-as-you-go model means users have to pay different amounts depending on how they use the AI. Subscription tiers are a little more simple, but these plans have their own terms and prices, which vary by company and model.
To clarify things, here’s what you need to know about the pricing models of the three most powerful models in the AI industry.
fable 5
First up is Anthropic’s latest release, the legendary Fable 5.
Subscribers to Claude Max, Pro, Team, and seat-based Enterprise plans can use Fable 5 with their plan’s existing token allocation until June 23rd. Starting from that date, the company will revert to a pay-as-you-go model for all Fable 5 users. That is, the more intensively the model is used, the more the customer has to pay.No matter what subscription tier you subscribe to.
According to a blog post published earlier this week, Anthropic intends to reinstate regular subscription-based token allocations for Fable “if sufficient capacity allows.” It is not yet clear what will happen to paid Claude subscribers who have not used up their entire token allocation by the June 23 deadline. We’ve reached out to the company for answers and will update this article as we learn more.
The important thing to remember here is that Fable 5 consumes more tokens than Anthropic’s previous models. So if you’re currently paying $100 a month for the Max 5x plan, you’ll continue to pay the same amount with Fable, but there’s a good chance you’ll hit your token limit sooner.
Starting on the 23rd, all users will have to pay $10 per million input tokens and $50 per million output tokens to use Fable.
According to common arithmetic shortcuts, one token translates into approximately 4 characters of text. Therefore, to earn 1 million tokens, you will need a huge amount of prompts. So if you just use Fable to write work emails or create recipes for dinner, for example, you can get good value from $10. Again, if that’s all you need for AI, you’re better off using a free chatbot. Using Fable to respond to a simple text chat is like driving your McLaren W1 to your neighbor’s house.
Fable 5 is specialized for long-running autonomous tasks, such as writing software code. This will require more tokens. Both input and output require hundreds of thousands to millions of tokens. Therefore, your monthly bill will be significantly higher than if you just input simple text prompts into the model. But if you’re already paying $200 a month for the Max 20x plan, you might not be paying more for usage credits than you are now. With 10 million input tokens and 5 million output tokens, your bill would be $350 (($10 x 10) + ($50 x 5)).
This means that the cost of using Fable 5 depends entirely on the demand of the task using that model. This is the basis of the pay-as-you-go model. If you tend to rely on models for complex tasks that require many steps and a long time, proceed with caution.
GPT-5.5 Pro
GPT-5.5 Pro, released in April, is the latest model to power ChatGPT. It’s available on OpenAI’s Pro plan ($200 per month), as well as the company’s Business ($30 per user per month) and Enterprise (custom pricing) tiers.
Meanwhile, developers using GPT-5.5 through OpenAI’s API will be billed on a pay-as-you-go model similar to the one that will begin rolling out to Fable later this month. At $5 per million input tokens and $30 per million output tokens, it’s significantly cheaper than Fable (and slightly more expensive than Anthropic’s second most valuable public model, Opus 4.8). It also comes with a batch tokenization option that’s 50% cheaper, essentially making OpenAI Servers can handle bundles of similar requests in a single “batch”, which increases computational efficiency, but also slows down response times.
gemini 3.5 flash
Google highlighted its unique blend of speed and agent capabilities with 3.5 Flash, the most powerful version of Gemini released last month.
It is available for free with limited usage and allows developers to build APIs for $1.50 per million input tokens and $9 per million output tokens. This is by far the most affordable option of the three models we’ve reviewed so far.
conclusion
Just as AI does not have a standardized pricing model across the industry, there is also wide variation in the advantages and disadvantages of each model.
For many users who just need a chatbot to act as a prestigious search engine, the free versions of Claude, ChatGPT, or Gemini are probably fine. If you have a job that requires more advanced models, such as for coding or research purposes, paying a subscription is probably the way to go. Pay attention to the fine print before making your selection, and be on the lookout for important phrases like “usage limits” and “pay-as-you-go.”
