Google Preps Skimaki to Reclaim Coding Crown. Flash Models Move Fast. So Does Irrelevance.
Google is reportedly preparing a new Gemini model codenamed Skimaki, potentially launching as early as this week. The model may outperform rivals from OpenAI and Anthropic in coding tasks specifically. Google has shipped Flash models rapidly, with Gemini 3.6 Flash in July followed by 3.7 Flash three weeks later, though the company has lost ground since briefly leading with Gemini 3.0 last November.
This illustrates the cadence advantage of smaller, specialized models. Flash models require far less compute than Pro variants, enabling faster iteration cycles. The mental model is simple: ship narrow, ship often. The mechanism is resource asymmetry. Lighter models let a company test improvements at a pace that flagship releases cannot match.
Google is developing the Skimaki model and has been rapidly shipping Flash variants, with 3.6 Flash in July and 3.7 Flash three weeks later. The company briefly led the AI race with Gemini 3.0 last November but has since lost ground to OpenAI and Anthropic.
- Open Google Gemini in your browser and test it with a coding task, such as asking it to write a Python function that sorts a list of dictionaries by a specific key.
- Take the same prompt to a competing model like ChatGPT or Claude and compare the output quality, speed, and accuracy.
- Note the differences in how each model handles edge cases or follow-up corrections, which is the exact type of coding benchmark comparison that Skimaki is being built to win.