Alphabet CEO Sundar Pichai said on the July 22 Q2 earnings call that coding and agentic coding are areas Google needs to improve. Google says Gemini 4 has entered its most ambitious pretraining run, and Pichai said the larger base model will be needed to compete at the next frontier. No release date, API specifications or independently verified coding results are available.
Alphabet CEO Sundar Pichai said during the company's July 22 earnings call that coding and agentic coding are areas Google needs to improve. He paired that acknowledgment with a commitment to keep iterating on the Gemini 3.6 Flash line and to train a larger Gemini 4 base model.
Google's edited account of Pichai's prepared remarks says Gemini 4 has entered the company's most ambitious pretraining run yet. In the question-and-answer portion reproduced by Investing.com, Pichai said Google will need Gemini 4 to compete at the level where the frontier is expected to be when the model is released.
Current models and the next training run
Google describes Gemini 3.6 Flash as its workhorse model and says it is testing the model with customers for coding. Pichai said the company expects continued iterations rather than a single step change. Search Engine Journal reports that Gemini 3.5 Pro remains in partner testing after missing an earlier expected release window.
Gemini 4 is further out. Google has disclosed that pretraining is underway, but it has not provided a release date, parameter count, context window, pricing, API availability or a reproducible set of coding evaluations. Pichai's confidence in internal progress is therefore a roadmap statement, not independent evidence of the model's eventual capability.
What engineering teams can evaluate
Coding-agent quality depends on more than generating plausible code. Teams need to measure repository navigation, tool use, test execution, error recovery, permission boundaries and performance on long-running tasks. A vendor's general model strength or aggregate token usage does not establish reliability on those workflows.
Google reports that more than 9 million developers use its models each month and that its model APIs process about 22 billion tokens per minute. Those figures demonstrate scale, but they do not answer how often coding agents complete representative tasks safely and correctly.
The near-term signal is Google's explicit prioritization of coding, not a finished Gemini 4 product. Technical buyers should wait for released APIs and compare them on their own repositories, test suites, latency budgets and failure-handling requirements before treating the larger training run as a capability gain.
Key Points #
- 1Pichai said coding and agentic coding are areas where Google needs to improve.
- 2Google says Gemini 4 is in its most ambitious pretraining run, with no release date or public API details.
- 3Developer adoption and token volume show scale but do not establish coding-agent reliability.
Scoring Rationale #
Google's explicit acknowledgment that coding and agentic coding need improvement is relevant to teams comparing frontier model providers. Gemini 4 remains an unreleased pretraining effort, so the event is a significant roadmap signal rather than a deployable capability.
Sources #
Primary source and supporting public references used for this report.
Practice with real Ad Tech data
90 SQL & Python problems · 15 industry datasets
[Active Search Campaigns by BudgetEasy](/problems/sql/active-search-campaigns-by-budget)
[High CPC Clicks & Poor Landing PagesMedium](/problems/sql/high-cpc-clicks-poor-landing-page)
[Campaign ROAS by Attribution ModelHard](/problems/sql/campaign-roas-by-attribution-model)
250 free problems · No credit card