Gemini 4 Hydrogen vs Argon, Opus 5.5 and Sol: same prompts, the real bill
Google says its new fast model, Gemini 4 Hydrogen, scores above Gemini 4 Argon and costs a quarter less per token. So it got the exact six prompts (a life animation, a voxel castle, a short film, a 3D car showroom, a driving game, a music studio) that Gemini 4 Argon, Claude Opus 5.5 and GPT-6.1 Sol built earlier, one request each, word for word. Everything runs in your browser.
A whole life, drawn in code
A human life from first heartbeat to old age, every frame drawn by JavaScript on one canvas, with a score synthesised in the browser.
A sixty-second paper-cut film: cradle, blackboard, late nights at a desk, autumn, an armchair by the window. Ran first time, no edits.
Gemini 4 Argon (reasoning low), from the video-27 run. No edits.
Opus 5.5, from the video-24 run. No edits.
GPT-6.1 Sol, from the video-24 run. No edits.
A voxel castle school on a lake
Thousands of instanced blocks, a day and night cycle and a fly camera.
Towers, a bridge, a harbour and a greenhouse. It opens at golden hour and very dark: move the time slider or start the cinematic fly-around. Ran first time, no edits.
Gemini 4 Argon (reasoning low), from the video-27 run. One marked text edit to its subtitle line (see the video-27 page).
Opus 5.5, from the video-24 run. No edits.
GPT-6.1 Sol, from the video-24 run. No edits.
A short film written in code
The Last Lighthouse: a storm, a sweeping beam, a boat fighting for the harbour, dawn. Sound is synthesised too.
With ONE marked line added: the depth-of-field pass received a copy of the render target texture, which three.js cannot clone, so every frame was black. After the fix: rain, a storm sea, the sweeping beam, a red boat.
Exactly as the model wrote it (text up to its first closing html tag; it then wrote a second copy of the file). Title cards and music play, the picture stays black.
Gemini 4 Argon (reasoning low), from the video-27 run. One marked fix; its text cards render as black boxes.
Opus 5.5, from the video-24 run, with that run's one marked fix.
GPT-6.1 Sol, from the video-24 run. No edits.
A 3D car showroom
An original electric crossover modelled entirely in code: paint, doors, lights, wheels, a night city.
Every control works (paint, four doors and a tailgate, three wheel styles, lights, turntable, night city), but the body panels do not fit together: the front sits apart and the bonnet is missing. Ran first time, no edits.
Gemini 4 Argon (reasoning low), from the video-27 run. No edits.
Opus 5.5, from the video-24 run. No edits.
GPT-6.1 Sol, from the video-24 run. No edits.
A sunset coast drive
A stylised driving game with a convertible, traffic, a speedometer and engine sound.
A red convertible on a cliff road at sunset, a round speedometer and a Wrecked screen with distance and top speed. Ran first time, no edits.
Gemini 4 Argon (reasoning low), from the video-27 run. No edits; the restart key throws an error, reload to play again.
Opus 5.5, from the video-24 run, with that run's one marked fix.
GPT-6.1 Sol, from the video-24 run. No edits.
A music studio in the browser
A step sequencer, synths, drums, effects and a button that composes a song.
Seven tracks, knobs on each, a live oscilloscope and spectrum, three moods and Generate song. Ran first time, no edits.
Gemini 4 Argon (reasoning low), from the video-27 run. No edits.
Opus 5.5, from the video-24 run. No edits.
GPT-6.1 Sol, from the video-24 run. No edits.
Bonus: the same prompts on xhigh
All six xhigh runs stopped at 65,536 output tokens, the limit of the OpenRouter route they were sent to ($0.49 each). Three had already written a whole file (kept up to its first closing html tag, nothing edited); the car, the film and the life animation were cut mid-file and are not published.
Opens in daylight; the nicer of Hydrogen's two castles.
A working studio with oscilloscope, spectrum and moods.
The menu shows, but in our test the game screen stayed black after starting.
What it cost
API cost of the six requests, as billed by OpenRouter. Hydrogen is $1.50 per million input tokens and $7.50 per million output tokens; Argon was $2 and $10 at the time of its run; Opus 5.5 is $4 and $20; Sol is $2 and $10. Hydrogen is cheaper per token but wrote about 53% more output tokens than Argon (233k vs 152k).
| Model | Six requests, total cost | Total generation time |
|---|---|---|
| Gemini 4 Hydrogen | $1.75 | 26.6 min |
| Gemini 4 Hydrogen · xhigh | $2.95 | 37.5 min |
| Gemini 4 Argon | $1.52 | 23.4 min |
| Claude Opus 5.5 | $2.70 | 20.8 min |
| GPT-6.1 Sol | $0.77 | 26.2 min |
The exact prompts are for Ruocco Academy members. The Academy is not open yet: join the waiting list and you get one email when it opens. The demos on this page are free to use for everyone.
Join the waiting list