We might be two months away from an affordable Fable-class model, and the wait is killing me

Anthropic made Claude Fable 5 available to the public on June 9. Access was interrupted almost immediately because of a government export-control issue, but it was restored globally on July 1. Fable 5 launched at $10 per million input tokens and $50 per million output tokens, which is still expensive for ordinary users. Anthropic’s launch announcement and its redeployment update explain the timeline.

What has me thinking about this is the idea that Chinese AI models may only be a few months behind the best American models. I keep hearing estimates of three to six months. I don’t know if anyone can measure that gap perfectly, but if the general trend is right, then we may be getting close to a model that performs around today’s Fable level without costing anything close to Fable prices.

That could happen sooner than people think. Six months after Fable’s original release is December 9. I’m not saying a Chinese company is guaranteed to release a perfect Fable copy on that exact date. I’m saying that if the capability gap really is only a few months, then we could be looking at a much cheaper version of this kind of model before the end of the year.

Opus 5.5 is already an example of the frontier moving downward. Anthropic says Opus 5.5 performs around the level of Fable 5.1 on many tasks, while its listed price is $4 per million input tokens and $20 per million output tokens. That is still not cheap by normal software standards, but it is a meaningful step down from Fable’s $10 and $50 prices. Anthropic’s Opus 5.5 announcement includes those comparisons and prices.

I think this is where the AI market starts changing for most people. Once a model reaches the level of current Opus or Fable-class systems, most everyday users probably will not need anything better. If someone is writing, researching, studying, running a small business, making videos, or doing ordinary knowledge work, that level of AI will probably be more than enough. My guess is that around 90% of users would be satisfied with it.

The people who still need something beyond that would mostly be working at the frontier of their fields. They might be doing advanced scientific research, building the next generation of AI systems, solving unusually difficult engineering problems, or running large autonomous workflows. For everyone else, the difference between the absolute best model and a model that is merely extremely good may not matter very much.

Think about a future where a Fable-class model costs what DeepSeek V4.1 Flash costs right now. DeepSeek currently lists Flash at $0.15 per million uncached input tokens and $0.60 per million output tokens during off-peak hours, with peak prices of $0.30 and $1.20. DeepSeek’s official pricing makes the target pretty clear.

That is what I want to see. At that price, people would not have so much anxiety about using powerful models. They would not have to feel price anxiety every time they send a difficult request or start a long task. They could try more ideas, run longer workflows, and use AI agents without constantly wondering how expensive the next prompt is going to be.

That is why the wait is killing me. I don’t necessarily need the absolute best model in the world. I want a model that is good enough to feel like Fable, but cheap enough that normal people can use it all day. If the three-to-six-month estimate is even close to being right, that model may be much closer than it seems.