Archive · 2 Sep 2026
Google Prepares Gemini 3.8 Flash to Challenge AI Rivals
Google is working on a new AI model called Gemini 3.8 Flash.
It is designed to compete with AI systems from OpenAI and Anthropic, especially for computer coding.
Reports say Google is using more reinforcement learning to train it.
This means the model practices tasks repeatedly and learns from feedback.
Google has also reportedly hired Barret Zoph, an AI researcher with experience at Thinking Machines Lab and OpenAI.
His work will focus on helping AI systems learn after their initial training.
If Gemini 3.8 Flash launches successfully, Google may become stronger in AI coding.
However, that would not necessarily mean Google is the overall leader in artificial intelligence.
Google is preparing its next Gemini model to compete with OpenAI and Anthropic in AI coding.
The Wall Street Journal reports that Google has increased its use of reinforcement learning.
Reinforcement learning allows AI models to improve through repeated attempts and feedback.
Google reportedly hired Barret Zoph as a vice president of research focused on reinforcement learning and post-training.
A successful Gemini 3.8 Flash launch could show progress in a fast-growing field without proving Google leads the overall AI race.
- Who
- Google, OpenAI, Anthropic, and Barret Zoph are identified in the report.
- What
- Google is preparing a next-generation Gemini model, reportedly Gemini 3.8 Flash, for AI coding competition.
- When
- The report does not specify a launch date.
- Where
- The report does not specify a location.
- Why
- Google is seeking to improve its position in AI coding, a rapidly growing area of the industry.
- Model
- Gemini 3.8 Flash
- Reported competitors
- OpenAI and Anthropic
- Training method
- Reinforcement learning
- Reinforcement learning
- A process in which models improve through repeated attempts and feedback
- New hire
- Barret Zoph, reportedly appointed vice president of research
- Research focus
- Reinforcement learning and post-training
- Broader significance
- A successful launch would indicate progress in AI coding but not necessarily overall industry leadership