Summary
Google has highlighted four developer projects made with Gemini 3.8 Flash, from an orbital-path simulation to an interactive transmission model. The company attributes the model’s gains over Gemini 3.7 Flash to extra reasoning steps and repeated tool calls on complex tasks.
Google highlighted four developer projects made with Gemini 3.8 Flash on September 28, 2026, including an orbital-path simulation, an animated ink-painting effect, a 3D dinosaur skeleton and an interactive automatic-transmission model. The company says the model improves on Gemini 3.7 Flash in software engineering, agentic tasks and multistep reasoning, and attributes the gains to taking extra reasoning steps and making iterative tool calls on complex tasks.
Four projects explore different capabilities
Ashutosh Shrivastava used Gemini 3.8 Flash with Google Antigravity to make a simulation mapping the live paths of satellites, space stations and orbital rockets. The post describes this as one example of developers using the model to build simulations.
Noctus used the model’s multimodal capabilities to animate Seigaiha waves, a traditional Japanese pattern, in a style resembling a moving ink painting. This example uses image-related generation rather than a conventional software-engineering task.
For a more structured challenge, Emily prompted Gemini to create a 3D Tyrannosaurus rex skeleton in four phases. The prompt included strict requirements, prohibited shortcuts and accuracy checks. Those checks formed part of the task; the post does not report a separate scientific assessment of the resulting skeleton.
Hakm asked the model to build an interactive automatic-transmission simulation from scratch. The prototype described by Google includes 10 camera views, four display modes, labels and a side panel that explains a component when selected.
Google’s account of the model’s approach
Google characterises Gemini 3.8 Flash as a “workhorse” model for complex reasoning and software engineering. It says the model often approaches the performance of higher-cost frontier models, and that on harder tasks it can work through more reasoning steps and call tools repeatedly to improve its output.
The four projects show a range of ways builders are applying the model, from visual creation to interactive prototypes. They are demonstrations rather than controlled comparisons of model performance: the announcement provides qualitative claims about gains over Gemini 3.7 Flash but no benchmark scores for the examples.