During Alphabet’s Q2 2026 earnings call, CEO Sundar Pichai emphasized the need for improvement in Google’s coding and agentic coding capabilities. Pichai stated that a larger Gemini 4 base model is necessary to maintain competitiveness in artificial intelligence. His comments followed the introduction of Gemini 3.6 Flash and the announcement that Gemini 4 is currently in pretraining.
Pichai expressed confidence in Google’s overall performance but acknowledged gaps in agentic coding development. He indicated that reaching the next breakthrough in AI depends on larger base models and said Google is currently training Gemini 4 to be competitive. In May, Pichai had noted on the Hard Fork podcast that Google was “a bit behind” in agentic coding due to the lack of a developer-facing product that generates usage data.
The release of Gemini 3.5 Pro remains delayed due to coding issues. The model is currently testing with partners, and Pichai stated it would be available “as soon as it’s ready.” Bloomberg reported that the performance issues with coding have contributed to the delay. A late June update to the model’s training data did not meet expectations.
Google recently launched Gemini 3.6 Flash, which produces 17% fewer output tokens compared to Gemini 3.5 Flash and offers enhanced coding capabilities. In addition, Gemini 3.5 Flash-Lite was released on July 21, targeting faster and lower-cost operations for high-volume workloads.
According to a DeepSWE coding benchmark, Gemini 3.6 Flash scored 49%, a 12% increase from the 37% score with 3.5 Flash. The timeline for the release of 3.5 Pro and the planned monthly release schedule are critical for Google’s product strategy, but no release date has yet been confirmed for Gemini 4, which Pichai described as the most ambitious pretraining process to date. Google faces further scrutiny as two senior AI researchers departed for OpenAI and Anthropic amid concerns regarding the company’s standing in AI coding tools.
