In this pixel-art animation comparison, the author reported that Gemini 4 Neon took 2 minutes longer and cost about half as much as Opus 5.5, whose output the author preferred overall. Original X post
What the comparison covers: The models' output for this pixel-art waterfront city animation and the time and cost the author recorded.
Limits of the comparison: The results cannot establish the speed, cost, or success rate of other coding or generation tasks.
Model versions covered: Gemini 4 Neon and Opus 5.5, as named in the original post. The post leaves the API model IDs and snapshots unspecified.
Test environment or client: Unspecified.
Reasoning levels and parameters: The author says both models ran at their respective highest available reasoning levels. The post leaves the level names and other parameters unspecified.
The author says both models received the same prompt and ran at their respective highest available reasoning levels. The post gives one time and one cost for each model and leaves the number of runs, the start and end points for timing, and the basis for billing unspecified. Original X post
The author thought Neon added a lot of detail and produced the best result he'd seen from the Gemini 4 family on this test. He still found the Opus 5.5 result clearly better overall. The post gives no scoring criteria. Original X post
| Model | Time reported by the author | Cost reported by the author |
|---|---|---|
| Gemini 4 Neon | 20 minutes | About 3.5 |
| Opus 5.5 | 18 minutes | About US$7.2 |
The original post omits the currency symbol for Neon. Because the post presents the two figures together, this summary treats them as amounts in the same currency. On that reading, 3.5 ÷ 7.2 is about 49%. The original post and visible replies contain no invoices or timing logs. Original X post
The attached 12-second GIF stacks two pixel-art waterfront city animations, with Gemini 4 Neon labeled at the top and Opus 5.5 at the bottom. Both scenes show boats moving and the water's surface changing over time. The GIF is a preview of the animations. Original X post and attached GIF
The author identified the shared prompt as a control for the comparison. The post leaves unspecified how much compute each model used at its highest reasoning level. The reported times and costs apply only to this comparison, and the assessment of visual quality reflects the author's own judgment. Original X post
The original post and visible replies do not include the full prompt, runnable source code, or run configuration. The available material lets readers check what the author reported and view the preview. Rerunning the comparison would require the full input, model versions, reasoning settings, and the methods used to measure time and cost. Original X post
Gemini 4 Neon