Google is reportedly preparing a new Gemini model focused heavily on coding, with internal tests suggesting a much stronger challenge to OpenAI and Anthropic.
Google appears to be preparing another major move in the AI race, and this time the target is clear: coding. The company’s AI research division is reportedly getting ready to release Gemini 3.8 Flash, an unreleased model that has been tested internally and is said to deliver a significant improvement in software development tasks. The model is reportedly known inside Google as “Skimaki”, and an external release could arrive as early as this week.
What makes the story interesting is not simply another Gemini version number. Internal testing reportedly suggests that Google’s engineers have preferred the new model over Anthropic’s Claude Opus in some coding evaluations. If those results translate into real-world performance after release, Google could be preparing one of its most important Gemini upgrades yet.
Google Is Going After Coding
Coding has become one of the most important battlegrounds in generative AI.
Models from OpenAI and Anthropic have built a strong reputation among developers for writing, debugging and modifying code, while Google has increasingly pushed Gemini into the same territory. Gemini 3.8 Flash appears to be part of that effort.
According to reporting based on people familiar with the project, Google engineers have been testing the model through Jetski, an internal coding environment used by the company. In head-to-head testing, some engineers reportedly preferred Gemini 3.8 Flash to Anthropic’s Opus model.
That is still an internal result, not an independent benchmark. Google has not publicly released the model or published a complete set of performance figures, so it would be premature to declare a new winner in AI coding.
But the direction is significant.
Instead of relying only on increasingly massive models, Google is continuing to develop its Flash family around speed, efficiency and the ability to handle large volumes of AI work without the enormous computing requirements associated with the company’s larger models.
A Smaller Model With a Bigger Job
The Flash name has become important to Google’s strategy.
Google has been using the Flash line to deliver models that are faster and generally more economical to run than its most powerful versions. That makes them particularly interesting for coding assistants and AI agents, where a system may need to make hundreds of model calls while working through a complicated task.
For developers, that can matter just as much as raw intelligence.
An AI model that can write good code but takes too long or costs too much to operate becomes difficult to use inside large development workflows. A faster model that can repeatedly inspect files, write code, run tests and correct mistakes can be far more useful in practice.
That is the market Google appears to be targeting with Gemini 3.8 Flash.
The company has not yet publicly confirmed the model’s specifications, pricing or final release date. Reports currently point to a possible launch as soon as this week, but that remains unconfirmed until Google announces it.
The Timing Is Important
Google’s timing is particularly interesting because the AI coding race has become much more competitive.
OpenAI and Anthropic have both pushed their models toward increasingly autonomous software-development workflows, where an AI system is expected to do more than simply generate a few lines of code. Modern coding agents can inspect repositories, understand existing projects, modify multiple files, run tests and iterate on their own.
Google wants Gemini to compete in exactly that environment.
The company already has Gemini integrated across its developer ecosystem, while its internal teams are using AI heavily in their own software workflows. Improving coding performance therefore has implications far beyond a chatbot benchmark.
It could influence Google’s developer tools, cloud business and the wider Gemini ecosystem.
And there is another reason the release matters: Google cannot afford to move slowly.
The company has produced several strong AI models, but the market changes rapidly enough that a lead can disappear within weeks. Internal testing of a new Flash model suggests Google is trying to shorten that cycle and get useful improvements into developers’ hands faster.
Google Has Not Won Yet
There is an important distinction between what is known and what is being reported.
Gemini 3.8 Flash has not been officially launched, and Google has not yet published independent benchmark results proving that it beats OpenAI or Anthropic’s best coding models.
The strongest claims so far come from internal testing reported by people familiar with the model. That makes the story promising, but not conclusive.
Google also has to prove that performance inside its own tools translates into real-world results for outside developers.
If Gemini 3.8 Flash can combine strong coding ability with the speed and efficiency expected from the Flash family, however, it could become a particularly important weapon for Google’s next phase of the AI race.
For now, the most interesting part is that Google appears to be preparing to make its move sooner rather than later.
And developers may soon get the chance to find out whether “Skimaki” is actually as good as Google’s internal testing suggests.
