As artificial intelligence models continue to advance, the AI race is heating up with Anthropic and OpenAI announcing their latest powerful AI models. Amid the increasing competition, Google DeepMind could launch a new AI model that may improve its coding abilities, according to a report from The Wall Street Journal.
According to this report, the upcoming Google model could be called Gemini 3.8 Flash. It is reportedly known as Skimaki internally. Here is everything we know about the alleged AI model.
Gemini 3.8 Flash Reportedly Impresses Google Engineers
Reportedly, Google engineers preferred the new model over Anthropic’s Opus model, used for coding tasks. If this model also performs well on widely used AI benchmarks, it may help Google address concerns about its position in the increasingly competitive AI market. However, the new Flash model is not likely to replace Google’s most powerful AI models.
According to WSJ, Google has particularly increased its use of reinforcement learning. This part of AI training allows models to improve their abilities through repeated attempts and feedback. Google has also reportedly hired Barret Zoph, a former Thinking Machines Lab co-founder and OpenAI post-training leader, as a vice president of research. His work is focused on reinforcement learning and post-training. A successful Gemini 3.8 Flash launch would not necessarily mean that Google has regained the lead in the overall AI model race. Instead, it could show that the company is making progress in one of the industry’s fastest-growing areas.
The biggest question now is whether Gemini 3.8 Flash can match the internal results reported by Google employees when it becomes publicly available. Strong coding performance could give Google a much-needed boost as developers increasingly use AI to build and maintain software. But the broader AI race will depend on more than coding alone. Google will still need to demonstrate that its larger models can compete with the latest systems from OpenAI and Anthropic across a wider range of tasks.
