Games and apps written by the latest AI models, each one tested live in a @VitoMind video. Pick a collection, open any demo and try it yourself in the browser.
Seven prompts on GPT-6.2 Astra and Sol, two on Claude Fable 5.2. Play every build, see time, tokens and the bill.
Claude Opus 5.5 turned one long video into 5 Shorts, a newsletter, LinkedIn posts, an X thread and a 4-week calendar.
Qwen 4 Max and Qwen 3.8 Max on the six prompts Opus 5.5, Sol and Gemini built. Play every build, see the bill.
Gemini 4 Carbon got the six prompts Argon, Hydrogen, Opus 5.5 and Sol built. Play every build, see the bill.
Gemini 4 Hydrogen got the six prompts Argon, Opus 5.5 and Sol already built. Play every build and see the real bill.
One FPS prompt and one lake prompt, sent once to Claude Fable 5.2, Meta Muse Ember and the free MiniMax M3.1 Flash.
Three viral-format 3D Shorts written by Claude Opus 5.5 as three.js code: glass, oxygen and your skeleton vanish for 5 seconds.
Ripple-65B Preview and Tencent Hy4 on the exact six prompts Opus 5.5, GPT-6.1 Sol and Kimi K3.1 Pro got. Play all 30 builds.
Inside Herdr, Claude Code started Codex, Sonnet, Gemini and Grok and gave each one scene of an ASCII film. Watch the four scenes.
Claude Code leading Codex and Sonnet through Herdr vs Claude Opus alone: two races, four builds you can play.
Three real-life Claude ideas, one prompt each to Opus 5.5: a 3D room planner, a drone lab and a smart-home wall panel. No fixes.
GPT-6 Astra and Claude Opus 5.5 each wrote all 1,210 weights of a tiny digit-reading network by hand, with no training and no examples. Draw a digit and test them live.
Two design skills for coding agents on the same prompt and the same model, Claude Opus 5.5, plus a run with no skill. Open all three pages.
Two 2B open models on a 12 GB RTX 3060: very fast, perfect at extracting data, and what they wrote for code, exactly as written.
Qwen3.8 27B on Kaggle's free T4 x2 and on a 12 GB RTX 3060: same prompts, real times, the builds as written and with marked fixes.
Kling 4.0, Seedance 2.5, MiniMax H3, Wan 3.0 Prime, Grok Imagine 1.5, FLUX 3 Video and Veo 3.1 on one prompt. Watch the clips, vote, reveal.
Does more thinking pay? Claude Opus 5.5 at effort low, high and max vs Gemini 4 Argon on three builds, with the real bill.
A 125B model on a 12 GB RTX 3060, on the exact prompts Claude Opus 5.5 and GPT-6.1 Sol got: 6 builds, real time, the honest result.
GPT-6.1 Terra vs Claude Opus 5.5 vs Haiku 5.5: same four prompts, 12 builds you can use in the browser.
Kimi K3.1 Pro on the exact prompts Claude Opus 5.5, GPT-6.1 Sol and Fable 5.1 got: 8 builds, real cost and time. Play all 25.
Two local models at full context in OpenCode on a 12 GB RTX 3060, same project prompt. Play all builds.
Recursive Flint on an RTX 3060 vs Gemini 4 Argon and Opus 5.5: same four prompts, 12 builds plus a live Snake.
Two open models running offline on a 12 GB RTX 3060, same game prompt. Speeds we measured, and both games to play.
Two models OpenCode serves for free, same three prompts: a drift racer, a launch page, a motion film. Built for $0. Try all six.
Three motion videos made for $0 by DeepSeek V4.1 Flash Free (OpenCode) inside OpenDesign Motion. Watch the MP4s, open the live code.
Three prompts, four models: the new DeepSeek V4.1 Flash-1017, the old V4.1 Flash, Claude Opus 5.5 and GPT-6.1 Sol. Voxel siege, windy hill, 3D headphones. Play all 12.
The tiny Japanese voxel sandbox prompt behind the Claude Next clip, given to Claude Fable 5.1, Claude Opus 5.5 and GLM 5.3 Flash. Plus floating sky islands. Play all six builds in your browser.
Meta's open Muse Spark Open against Inkling, HY4 Preview, Step-5 Preview and Nemotron on one survivors game, and against Muse Wonder on a voxel pagoda and a voxel city you can blow up.
Four original browser demos generated with Claude Opus 5.5 and Ling 3.1 Flash: an island game, sequencer, six-style animation and short film.
Three free models in OpenCode (Cat Jam, MiMo-V2.6-Flash, Big Pickle) got the same three prompts: a pixel-art studio, a 3D glider game and the Fable 5.5 civilization film challenge. Play all nine builds in your browser.
Three free OpenRouter models and Claude Haiku 5.5 got the same six creative-code prompts: a life animation, a voxel castle, a short film, a 3D car showroom, a driving game and a music studio. Play all the builds in your browser.
Google's Gemini 4 Argon, Claude Opus 5.5 and GPT-6.1 Sol got the same six creative-code prompts: a life animation, a voxel castle, a short film, a 3D car showroom, a driving game and a music studio. Play all eighteen builds in your browser.
Meta's Muse Wonder, Claude Opus 5.5 and GPT-6.1 Sol got the same six creative-code prompts, plus three agent-style tests: a hidden SQL bug, a coding agent with a test suite and a translated newspaper headline. Play all eighteen builds in your browser.
Claude Opus 5.5 and GPT-6.1 Sol got plain behavior notes for three games: a brick breaker with stereo sound, a bullet-hell shooter and a voxel giant game. Play all six in your browser.
Is Gemini's Flash really at Opus level? The same three prompts to both: a floatplane simulator, a village of NPCs that remember you and a swarm of bug hunters. Play all six in your browser.
Six creative-code builds, same prompts, given to Claude Opus 5.5 and GPT-6.1 Sol: a life animation, a voxel castle, a short film, a 3D car showroom, a driving game and a music studio. Play all twelve in your browser.
How close is the best Chinese open model to the best you can buy? The same four prompts to Claude Fable 5.1 and Kimi K3: a game controller in SVG, a tactical shooter, a lit roguelike and a spreadsheet app. Play all eight in your browser.
Google's Playground launch prompt (a cozy ube latte game) built in three chat messages with Gemini. Play all three versions in your browser.
Six original browser demos: Step-5 and GPT-6.1 racers and voxel gardens, a working CSV dashboard and an editable animation studio. Tested for VitoMind.
A multiplayer voxel sandbox written by Opus 5.5, GPT-6.1 Sol, Luna, Haiku 5.5 and Step-5. Play them and vote.
Haiku 5.5 and GPT-6 Luna, same price per token, on the prompts Opus 5.5, Sol and Sonnet 5.5 got. 19 demos in your browser.
Mistral's 1-trillion-parameter open model on the same four prompts as Opus 5.5 and GPT-6.1 Sol. All 12 demos run in your browser.
The same four prompts to the models behind ChatGPT Pro and Claude Max: a racing game, a working 3D kitchen, a voxel pagoda and a card machine. Use them in your browser.
Germany's open Kolibri-1 against the models it says it beats: one game and one app, same prompts, ten demos you can use in the browser.
Four tests people ran on the leaked Claude Fable 5.5, given to Fable 5.1 and to the free Ling 3.1 Flash: an art-history film, a forest jeep, a rope puzzle and a controller designer. Use them all in your browser.
Demos written by Space Bunny Alpha (the free model the community links to MiniMax M3.1 Flash) and by MiniMax M3, on the same prompts: a voxel island, a wave shooter and a desktop OS. Play them all.
Five apps and games written in one prompt each by Claude Sonnet 5.5, Anthropic's mid-size model. Open them and try them live.
The same tram-game prompt given to GPT-6 Astra, an early Gemini 4 checkpoint, Claude Sonnet 5.5, GPT-6.1 Sol and Gemini 3.8 Flash. Play all five.
Four apps and games written in one prompt each by OpenAI's GPT-6.1 Sol, plus a head-to-head against Claude Sonnet 5.5 on the exact same prompts. Open them and try them live.
Six apps and games written in one prompt each by Space Bunny Alpha, the free mystery AI model. Open them and try them live.