Gemini 4 enters Google's frontier race

Gemini 4 looks less like an imminent launch than Google’s high-stakes reset for frontier AI. The real test will be whether a larger base model can close Google’s acknowledged gap in coding and agentic workflows.

• –Pre-training is confirmed, while most rumored capabilities remain speculation
• –Google is positioning Gemini 4 beyond incremental Flash-series updates
• –Developers should not plan against a Gemini 4 API or model ID yet
• –Coding, tool use, reasoning, and agent reliability will likely define its impact
• –The delayed Gemini 3.5 Pro makes Gemini 4’s eventual timing harder to predict