Fable 5 Extended to July 12: Why It Happened and What to Build
Prefer to watch? Watch on YouTube · Subscribe for new builds · Read the transcript
In this video I explain why Fable 5 access in Claude subscription plans got extended to July 12, and what I think you should build with the extra days. It's for anyone using Claude Code (or thinking about it) who wants to get the most out of Fable 5 before it moves to API pricing. I recorded this in July 2026, and these tools change fast, so check the current Anthropic docs for the latest on plans and model access.
What I cover in this video
- 0:00 Fable 5 access is extended to midnight July 12 at no extra cost on your subscription
- 0:45 Polymarket odds on OpenAI's next GPT release, and why I think that's behind the extension
- 1:32 What I'm building: Beckon, a voice-controlled ADE built with Fable 5
- 2:17 How Fable 5 compared to Opus 4.8 for me, plus a live voice command spinning up Claude Code agents
- 3:02 Hitting 100% usage, buying a second Max plan, and my other projects like ShipBench
- 3:48 Beckon's memory, workbench, agent loops, and local voice option
- 4:34 Why voice beats typing, and why this is the window to build your app idea
- 5:19 The first thing to do with any new model: refactor your codebase
Why Fable 5 got extended
The Fable 5 promotion started July 1. Everyone expected it to be pulled from subscriptions at midnight the night I recorded this, but Anthropic extended it to midnight July 12. Until then, Fable 5 is included in your subscription at no extra cost. After that, it moves to API pricing.
I think I know why. On Polymarket, the market for OpenAI's next GPT release had an 83% chance on July 9, which fits OpenAI's habit of releasing new models on a Thursday. OpenAI has said those models will be included in subscription plans, and early benchmarks on Terminal Bench 2.1 show them beating Fable 5 in plenty of tests. If Anthropic had pulled Fable 5 right before that launch, OpenAI would have had an easy shot at Anthropic's users. The extension pushes that back. Polymarket also had an 89% chance of a release by July 13, so OpenAI may just wait until the day after Fable 5's window closes.
Models are trading places back and forth now, and each round gets a little better. If you want to weigh the two ecosystems, my Claude Code vs Codex page is a good place to start.
What I'm building with Fable 5
I've been working on Beckon, a voice-controlled ADE. It connects to tools through MCP connectors, has vision built in, and is built on the Open Knowledge Format that Google released a few weeks before I recorded this. It includes:
- Rules, custom commands, recipes, and loops you can set up yourself
- A memory module and a workbench that tracks tasks
- Agent loops, with a view of which agent did what
- Voice control through GPT Realtime 2, plus a local Parakeet model if you want voice control without going to the cloud
- Different backgrounds you can switch between
I'm dogfooding it, using Beckon to build Beckon, shipping new features and then testing and iterating. I worked on it for a few weeks with Opus first. Once Fable 5 came out, I shipped much faster with much cleaner code and far fewer bugs. Opus 4.8 had a hard time with that, even on high or extra high.
In the video I give it a voice command live:
Hey Beckon, open three more Claude Code agents and tell one of them to give me a description of Fable 5 and its new features.
It opens the Claude Code agents and passes the request along, no typing involved.
Using every bit of the extension
I'd already hit 100% of my usage, which resets the next day. Because of the extension, I'm buying a second $200 Max plan so I can run loops and build other projects at full capacity. Beckon isn't the only thing on my list. I'm also building ShipBench, a benchmark tool that lets me test each new model myself on tasks from my daily workflow. I'm also doing some work for Google and Idea Browser.
My bigger point: this is a rare chance to build your app idea, whatever it is, while Fable 5 is part of your subscription. And if you aren't using Wispr Flow or another voice dictation tool yet, try it. Typing is starting to feel like the slow way, and a voice-controlled agent lets you build faster. For more on getting started, see Build an App with AI.
The first thing to do with a new model
Whenever I get access to a new model, whether it's Fable 5 or the next GPT, I have it go through my codebase first:
- Look over the entire codebase and refactor it
- Point out what it can improve and what changes to make
- Squash any bugs
- Look for security vulnerabilities
It's a simple habit that gets you real value from a stronger model right away, before you start on new features.
Key takeaways
- Fable 5 is in Claude subscriptions at no extra cost until midnight July 12. After that it moves to API pricing.
- The extension probably lines up with OpenAI's expected GPT release, which Polymarket put at 83% for July 9 and 89% for July 13.
- Fable 5 let me ship Beckon faster and with fewer bugs than Opus 4.8 did.
- Voice control, with tools like Wispr Flow or a voice-driven ADE like Beckon, can speed up how you build.
- When a new model drops, start by having it refactor your codebase, fix bugs, and check for security vulnerabilities.
FAQ
How long is Fable 5 included in Claude subscription plans?
The Fable 5 promotion started July 1 and was extended to midnight July 12. After that, it moves to API pricing.
Why did Anthropic extend Fable 5 access?
My guess is it's tied to OpenAI's next GPT release, which Polymarket put at an 83% chance by July 9. Keeping Fable 5 in subscriptions makes it harder for OpenAI to pull Anthropic users away.
What should I do first with Fable 5 or a new GPT model?
Have it review your whole codebase and refactor it, suggest improvements, squash bugs, and look for security vulnerabilities. I do this every time a new model comes out.
What is Beckon?
Beckon is a voice-controlled ADE I'm building with Fable 5. It has MCP connectors, built-in vision, memory, a task workbench, and agent loops. Voice runs through GPT Realtime 2, or a local Parakeet model if you want to stay off the cloud.
Transcript
Read the full transcript
0:00In a massive turn of events, you now have at Fable 5 access until July 12th midnight. Previously, this model was expected to be ripped from our hands and our subscriptions at midnight tonight, but they went ahead and approved it till July 12th. So, you have now Fable 5 access at no extra cost as part of your subscription plan. I'll put the link to this source down below. But, yeah, the promotion started July 1st and then it ends now July 12th. And I think I know why. Let's take a look. Look here on Polymarket, there is a GPT released by Trend that is actually said to July 9th 83%, which is pretty much in line when a
0:45new GPT model releases, which is usually on a Thursday. And that is Soul, Luna, and Terra. And these models bench out beautifully, really strong performers, and we have heard from OpenAI before say that that is going to be included inside of your subscription plan. So, when you see Fable 5 be extended on July 12th, we were all wondering what was Anthropic going to do, and now you have it. So, now OpenAI, their opportunity to steal market share from Anthropic users has now been delayed. So, I'm wondering if they are going to now release their models on July 13th, which is an 89% chance. We are in this weird era of this AI space where now models are just going back and forth. They're getting progressively better with Soul, Luna, and Terra and Soul Ultra beating Fable 5
1:32in a lot of the early benchmarks on Terminal Bench 2.1. So, this is interesting, but I just wanted to make this video and go into the detail of what you should be building. Let's get into it. So, I've been working on Hey Beacon, which is basically a voice-controlled ADE that has connectors via MCP, and it also has vision built in, built on the Open Knowledge Format released by Google just a few weeks ago. So, we have built-in rules, so you can set rules and custom commands, recipes, and loops. All this built out with Fable 5. I've been actually using the platform to build itself app and shipping new features and slowly dogfooding it and updating, testing, and iterating. And
2:17like I said, this was all made possible with Fable 5. I have been working on this for the last few weeks using Opus, but once Fable 5 came out, I was able to ship much faster, much cleaner code without introducing a ton of bugs, which Opus 4.8, even on high or extra high, had a hard time doing. So, like I said, this thing has all been made possible. Hey Becca, open three more Claude code agents and tell one of them to give me a description of Fable 5 and its new features. >> All right, let me set those up and pass along your request. >> So, >> Opening three Claude agents and sending the request to the new one. >> So, if you look here and just look at my usage, I have hit the 100% usage that
3:02resets tomorrow. And with this new news, I'm going to be buying a second Claude code plan. Another max plan for $200, that way I can use this to its full capability with running loops and building out other projects. This isn't the only thing that I'm building out. I'm also working on ShipBench, which is basically just a benchmark tool that when new models come out, I can then verify and then run the benchmarks myself based on tests that I do in my daily workflows. So, that's going to be another thing I build out. And I also do some work for Google and Idea Browser that I'm going to be building out as well, that's just going to help me manage my second brain. So, all this has memory included. You have your own
3:48memory module here. You have your own workbench where it tracks tasks down below, and everything is just organized beautiful for builders. I'll put the link to it down below. You can also run agent loops here, see what agents were doing what, and it can be really helpful because it's all controlled via GPT real time two via voice. We also use a local parakeet to be able to use as a local model. Say you wanted to, you know, not be in the cloud, but you actually wanted to use voice control and just have it stay local. Well, we have that built out as well. And you can also change different backgrounds. Whatever you want to do, trying to just build out an interface for builders that can be all controlled by voice. If you're not building with, you know, WhisperFlow or
4:34some type of voice dictation tool, I feel like typing is a thing in the past. So, being able to have a voice-controlled agent as an ADE to be able to build faster is the key. So, that's what I have working on right now, and I'm super fired up that Fable 5 has now been extended. I did want to just let you guys know that this is an opportunity of a lifetime to build out your app idea. Like, no matter what it is, take advantage of Fable 5 being in the subscription because it will be API pricing after this next few 5 days. So, with GPT 5.6 coming out here by July 10th, July 9th, there's an 80 to 90% chance that it is coming out according to Polymarket. Take that for what you
5:19will. But, there is a good indication that GPT, the new next model, will be coming out this Thursday, which is usually when they release their newest model drops. So, I'm just telling you to be ready. Take that idea. The first thing you should do when you're getting access to Fable 5 or these new GPT models is have it refactor your code base. Have it take a look at your entire code base, see what it can make better, see what changes it can make, squash any bugs, look for security vulnerabilities. Those are the things that I do when these new models drop. I just hope you're ready. It's such an exciting time to be alive, to be building out your idea. Check out Hey Beacon down below. We do have a waitlist as the beta drops this week. We'll see in the next video. Have a blessed day.
Ready to start building with AI?
Join Shipping Skool free — 4 live build sessions a week, courses, and a community that ships.
Join Shipping SkoolKeep reading
Claude Opus 5 vs Fable 5: Cost, Speed, and 3 Real Build Tests
I ran Claude Opus 5 and Fable 5 on high effort through three builds: a Three.js MMO, a Remotion video, and a landing page rebuild. Opus 5 won the game and the UI test, the video was close to a tie, and Opus cost about half as much. It's my new daily driver, but Fable 5 is much faster.
Fable 5 vs GPT-5.6 Benchmark: Horror Game, Blender, Remotion Tests
I ran Fable 5 and GPT-5.6, both on high effort, through three tests: a horror game, a Blender donut, and a Remotion video. GPT-5.6 was much faster and cheaper every time. Fable 5 made the better Blender donut, but its horror game froze, so for most people I think GPT-5.6 is the better value.
Fable 5 Review: Parallel Agents, Usage Limits, and Real Build Tests
I ran three Fable 5 agents in parallel at once. One redesigned my landing page in 16 minutes, one built a first-person RPG in 11, and one made a Remotion demo video in about 40. Fable 5 hallucinated less than Opus 4.8, but its usage limits make me unsure it's worth paying extra credits for.