You'll notice the change on your very next program. Until now, Smart Rabbit relied exclusively on Claude, Anthropic's AI. It was the obvious pick: Claude was the only mainstream AI able to display a truly interactive program right inside the conversation. But things have moved — ChatGPT can now do the same. So when you click « Create my program », the app now gives you a choice: Claude.ai or ChatGPT.

Same expert prompt, same interactive program on the way out, same recovery in the app. The difference? The model running behind the scenes, its free-tier limits, and what a paid plan can unlock. This article sorts it all out, with the real numbers as of September 2026.

🆕 What actually changes in the app

The flow is as simple as before, with one extra step in the middle:

1. You fill in your profile: age, level, physical condition, available equipment, goals, any limitations. Two minutes is enough, and everything stays stored on your device.
2. You click « Create my program » and a window asks your preference: Claude.ai or ChatGPT.
3. The app copies the expert prompt to your clipboard and opens the chosen AI in a new tab. Paste, send, done.
4. The AI builds your program as an interactive artifact: section tabs, day-by-day navigation, detailed sets and reps. You copy the generated code, paste it into the Session tab of Smart Rabbit, and your program displays in full — saved locally, available offline, no account needed.

One useful note: if you are in mainland China, where ChatGPT and Claude are not accessible, both options automatically redirect you to DeepSeek, which generates the same kind of interactive program.

🤖 Which model ACTUALLY runs on the free plans?

That is THE right question to ask in 2026, because models change every three months and marketing does not always spell out what the free tier may use. Between code names, « mini » and « pro » versions and usage windows, it is hard to keep up. Here is the exact state as of September 27, 2026, from official OpenAI and Anthropic sources.

Free ChatGPT: GPT-5.6 Luna, the unlimited one

Since August 6, 2026, the default model on free ChatGPT is GPT-5.6 Luna. OpenAI pairs it with an unbeatable argument for everyday users: unlimited text conversations. You can iterate on your program as much as you want — adjust a session, ask for an exercise variant, rephrase — without ever hitting a wall.

For questions that need deeper thinking, a « Think » button enables heavier reasoning, also powered by Luna. However, the most capable model in the family, GPT-5.6 Sol, is reserved for subscribers: it powers the Medium, High and Extra High reasoning levels on paid plans. And GPT-6 Astra, unveiled on September 3, 2026, is not available to the free plan at launch.

Free Claude: Sonnet 5, the most powerful but capped

Anthropic's philosophy is different: Claude Sonnet 5 became the default model on all Claude.ai accounts — free and Pro alike — on July 1, 2026, just one day after its release. Anthropic describes it as its most agentic mid-tier model, with performance approaching Opus 4.8 at a fraction of the cost. In other words: even on a free account, you are talking to a top-tier model, with no special button to press.

The trade-off is the cap: the free plan limits the number of messages per multi-hour window. Generating your Smart Rabbit program is well within range. But if you chain questions and adjustments in one long session, you will eventually hit the limit — and then you have to wait for the next window.

📊 The matchup at a glance

🥊 Claude or ChatGPT: which one for your program?

Honestly? Both do an excellent job with the Smart Rabbit prompt, precisely because the prompt does most of the expert work. But each profile has its natural candidate:

And there is a third option few people think of: use both. Generate your program with ChatGPT, then regenerate it with Claude, and compare. When two very different models converge on the same volumes, the same exercise choices and the same progression logic, you have confirmation the program is solid. It is free, it takes five minutes, and it is a great fitness literacy exercise.

💎 With a paid plan, the experience levels up

Free plans are more than enough to generate, regenerate and track your programs. But if you want to shift gears, here is what subscriptions unlock — all around $20/month:

• ChatGPT Plus: you get GPT-5.6 Sol, the strongest reasoner in the family, with Medium, High and Extra High levels. Concretely, for a workout program, it changes the game on complex situations: returning to training after an injury, combining lifting and running, training with minimal equipment or a chaotic schedule. Subscribers also get priority access to new releases, like GPT-6 in the Work and Codex tools.
• Claude Pro: it is the same Sonnet 5 as the free tier — no higher model for now — but with sharply raised message caps and wider usage windows. And Max plans open the door to Opus 5.5, Anthropic's most powerful model, for users who want the absolute maximum.

Our advice stays the same: start free. The Smart Rabbit prompt already encodes fifteen years of field expertise, and a program generated by free-tier GPT-5.6 Luna or Sonnet 5 will sit far above what most paid coaching apps produce. You can always upgrade later if you get hooked on iterating.

🧠 Why the prompt matters more than the model

This is the most important point of this article, so let us insist. Ask any AI « give me a workout program » and you will get a generic answer: three vague sessions, no real personalization, no physiological safety rails. The same model, driven by the Smart Rabbit prompt, produces something completely different, because the prompt embeds:

The model is the engine. The prompt is the coach holding the wheel. That is why Smart Rabbit stays relevant whichever AI you choose — and whichever one ships next month. When OpenAI releases GPT-6 for everyone, or Anthropic ships a Sonnet 6, your program automatically rides the upgrade: just click « Create my program » again.

🐰 Ready to generate your program with the AI of your choice?

🤝 So do you have to pick a side?

Absolutely not. That is the classic comparison trap: they present ChatGPT and Claude as rivals when they are complementary. The strategy we recommend:

The two AIs are different engines, but the wheel stays the same: your profile + the Smart Rabbit expert system. The final result lands in the same place — your Session tab, ready for your next workout.

🔮 A transparency note on models

AI models evolve every month, and the facts in this article reflect the situation as of September 27, 2026: GPT-5.6 Luna by default on free ChatGPT (since August 6), Sonnet 5 by default on Claude (since July 1), GPT-6 still reserved for paid OpenAI plans. Those names will change — maybe even before the end of the year. But the principle does not move: Smart Rabbit generates the expert prompt, the AI of your choice executes it, and you get a free interactive program in the app. Whatever the current fashion may be.

Conclusion

The news is simple and fits in one sentence: your free workout program can now be generated with ChatGPT as well as Claude. Free ChatGPT and its unlimited GPT-5.6 Luna to iterate without counting. Free Claude and its Sonnet 5 for the sharpest reasoning. And in both cases, the same expert prompt, the same interactive program, the same recovery in the app.

Now run the test. Generate your program with one, regenerate it with the other, compare volumes and progression. Then start your first session. The difference with a generic plan found on the internet, you will feel it from the very first reps.