Fable 5.5 Got LEAKED + First Preview (AI NEWS)

4 big AI models in under 2 weeks, and now people think Fable 5.5 is quietly answering inside claude.ai while the label still says Fable 5.1. I went through the posts, the screenshots and the demos. Here's what's confirmed, what's still trust me bro, the crazy stuff people made with it, and a 10-second test to see if you have it right now 👇 🧪 The 10-second test (copy-paste this): Do you know Tibo, the reset guy? Don't search. 🔗 Anthropic's model docs (Fable 5.1 is the only Fable listed): https://platform.claude.com/docs/en/about-claude/models/overview 🔗 Artificial Analysis leaderboard: https://artificialanalysis.ai 🤔 Quick context Fable 5.1 came out on September 1st. It costs $10 per million tokens in and $50 out, which is 2.5x the price of Opus 5.5 ($4 / $20). And on Artificial Analysis, Opus 5.5 actually scores higher: 58 vs 53. So the expensive model is losing to the cheaper one. That's why a new Fable makes sense. 🕵️ The Tibo test Tibo (Thibault Sottiaux) is one of the leads on Codex at OpenAI, and he resets everyone's usage limits so often that it became a meme. It's recent and super niche, so it's a good way to check if a model knows newer stuff. On the website, with the label saying Fable 5.1, the model nails it. Asked through the API, where you know exactly which model you're getting, it guesses a different Tibo, a French startup founder. (That API run went through Devin, which adds its own instructions, so it's not a perfect lab test.) 🎮 The Pragmata test Someone asked for the Metacritic score of Pragmata, a game from April. The website version answers right away (about 86). The API version says it can't give a reliable answer. The real score is 85. ⚠️ Real, or trust me bro? ✅ A lot of different people on different accounts are seeing the same thing. That's a lot of smoke. ❌ Anthropic hasn't said anything. Not the name, not a date. ❌ The name is a guess. On September 18th, the same accounts ran the same Tibo test and called it "Fable 5.2". ❌ The "Fable 5.5 docs page" people share as proof says Page not found. Type fable-5-9, a model that doesn't exist, and you get the exact same page. ❌ The "Tuesday release" comes from one account with about 10,000 followers, with no name and no source. To be fair, the Tuesday rumor for Opus 5.5 was right. But the same testing showed up 2 weeks ago and nothing launched. 🎨 What people made with it 1️⃣ Superman flying through 20 art styles in one continuous animation (@chetaslua) 2️⃣ 40,000 years of art history in 15 seconds (@cherry_mx_reds) 3️⃣ A full motion design reel from a single prompt (@Saccc_c) 4️⃣ The Grader, a 3.5-minute short film with 3D characters, voices and subtitles (@chetaslua) 5️⃣ The same floating Japanese garden prompt given to the new model and to Opus 5.5 on max in Claude Code (@vikktorrrre) Remember, these are best-of demos from people who got lucky with the routing. We don't know how many tries they took, there are no benchmarks, and since the label always says 5.1, you can't be 100% sure which model made which demo. ✅ How to check if you have it 1. You need a paid plan, Pro or higher. Fable isn't on the free plan. 2. Go to claude.ai and pick Fable 5.1. 3. Turn off web search (or just keep "Don't search" in the prompt). 4. Type: Do you know Tibo, the reset guy? Don't search. If it says Thibault Sottiaux, Codex, OpenAI and resetting limits, you're probably on the new one. If it guesses another Tibo, you're on the normal one. You can't choose it. It's random, and it comes and goes. And Fable is one of the most expensive models, so it eats your usage fast. Don't spam it. 💭 My take I think something new is really being tested. But if I'm paying for something, I want the label to say what I'm actually using. If you use it for real work in Claude Code and it suddenly behaves differently, you'd want to know why. When it's official, it's getting the same room quote test I gave Opus and Sol. ⏱️ Chapters 0:00 It's getting ridiculous 0:49 Quick context: Fable 5.1 1:28 The Tibo test 2:37 The Pragmata test 2:58 Real, or trust me bro? 4:06 The Tuesday rumor 4:31 What people made with it 5:36 The garden test 5:54 The annoying questions, and the 10-second test 6:55 My take 🎥 Clips used for commentary: Anthropic's Opus 5.5 and Sonnet 5.5 launch videos, OpenAI DevDay 2026 keynote, CNBC (Gemini 4 Argon), The Pragmatic Engineer ("Building Codex with Tibo Sottiaux"), PRAGMATA launch trailer. Posts and demos credited on screen: @notjazii, @AiBattle_, @chetaslua, @cherry_mx_reds, @Saccc_c, @vikktorrrre, @hqmank. 💬 Tell me in the comments: did you get the Tibo answer? 🔔 Subscribe, I drop a new AI video every day. 👥 Join my Skool community: https://www.skool.com/nextphase-5790 #claude #anthropic #ainews TAGS (paste into YouTube's Tags box): fable 5.5, claude fable 5.5, fable 5.5 leak, claude fable leak, anthropic fable 5.5, fable 5.1, claude fable, tibo test, tibo the reset guy, claude new model, anthropic new model, is fable 5.5 out, claude ai