Is Google losing the AI race? Sundar Pichai says no: just wait for Gemini 4

When thinking of an AI model, ChatGPT or perhaps Claude likely comes to mind. In contrast, Google’s Gemini is unlikely to top the list, as the company continues to strive to catch up with rivals in areas such as autonomous agent capabilities or programming tasks. Nevertheless, Google CEO Sundar Pichai maintains that the company is not lagging behind and assured investors that it will bounce back with Gemini 4.

During Alphabet’s (Google’s parent company) earnings call, Pichai refuted claims that Google was losing ground to competitors. Google’s top executive stated that “people will be satisfied” with the launch of Gemini 4, a model expected to represent a significant improvement over previous version.

Google’s CEO described Gemini 4 as a “very ambitious effort.” According to Sundar Pichai, Google’s new AI model is considerably larger than its predecessors, which should allow it to compete at the technological forefront. Pichai insisted that Google remains “strongly committed” and fully confident in its ability to lead the AI race.

It is worth noting that an AI model trained with a higher number of parameters is typically more powerful; in other words, the larger the model, the better its performance. While companies like Google and OpenAI do not disclose the parameters of their closed-weight AI models, Moonshot AI’s latest model, Kimi K3 (open-weight), was trained with over 2.8 trillion parameters—the highest figure for an open model. Pichai’s remarks follow online reports regarding the delay of Gemini 3.5 Pro. According to these reports, the AI model’s launch was postponed because it failed to meet Google’s internal benchmarks for coding tasks and autonomous agent capabilities—areas currently dominated by Anthropic and OpenAI.

Sundar Pichai acknowledges Google is trailing in AI coding capabilities

Responding to JPMorgan analyst Doug Anmuth, Pichai stated that Google still holds an advantage in certain respects. He noted, “There are many attributes where we remain at the forefront.” However, Google’s top executive acknowledged that the company had fallen behind in some key areas. “There are areas where we’ve admitted we need to improve, and coding—as well as coding agents—is one example of that,” he added.

To put this in context, coding has rapidly established itself as one of the most important capabilities for AI tools. Companies worldwide are increasingly using tools like Claude Code and Codex to write code. Even Meta and Elon Musk’s SpaceX have joined the AI coding race with new models.

Rather than focusing on the Gemini 3.5 Pro delay during the earnings call, Pichai repeatedly highlighted Gemini Flash—Google’s fastest and most cost-effective model—which is used in cybersecurity, customer service agents, data analysis, and enterprise software. He noted that Gemini 3.6 Flash, launched earlier this week, had improved by more than 10 points on a coding benchmark test compared to the previous version, while also using fewer tokens.

The company also introduced Gemini 3.5 Flash-Lite and a cybersecurity-focused Flash Cyber model, while keeping Gemini 3.5 Pro in the partner testing phase. Their defense reflected a broader effort to convince investors to look beyond a single delayed flagship model and consider Google’s AI ecosystem as a whole—including cloud infrastructure, custom chips, and Gemini products integrated across its various services.

Gemini surpasses 950 million users

This does not mean Google is failing to see a return on its investments. Sundar Pichai stated, “Our AI investments are redefining what is possible across all areas of our business,” adding that the company was seeing “promising momentum across the board.” According to company data, Gemini reached 950 million monthly active users, up from the 750 million recorded in the fourth quarter of 2025.

Is Google losing the AI race? Sundar Pichai says no: just wait for Gemini 4

Google CEO Sundar Pichai has dismissed claims that the company is falling behind in the AI race. Pichai asserts that Google will soon close the gap with the launch of Gemini 4.

When thinking of an AI model, ChatGPT or perhaps Claude likely comes to mind. In contrast, Google’s Gemini is unlikely to top the list, as the company continues to strive to catch up with rivals in areas such as autonomous agent capabilities or programming tasks. Nevertheless, Google CEO Sundar Pichai maintains that the company is not lagging behind and assured investors that it will bounce back with Gemini 4.

During Alphabet’s (Google’s parent company) earnings call, Pichai refuted claims that Google was losing ground to competitors. Google’s top executive stated that “people will be satisfied” with the launch of Gemini 4, a model expected to represent a significant improvement over previous version.

Google’s CEO described Gemini 4 as a “very ambitious effort.” According to Sundar Pichai, Google’s new AI model is considerably larger than its predecessors, which should allow it to compete at the technological forefront. Pichai insisted that Google remains “strongly committed” and fully confident in its ability to lead the AI race.

It is worth noting that an AI model trained with a higher number of parameters is typically more powerful; in other words, the larger the model, the better its performance. While companies like Google and OpenAI do not disclose the parameters of their closed-weight AI models, Moonshot AI’s latest model, Kimi K3 (open-weight), was trained with over 2.8 trillion parameters—the highest figure for an open model. Pichai’s remarks follow online reports regarding the delay of Gemini 3.5 Pro. According to these reports, the AI model’s launch was postponed because it failed to meet Google’s internal benchmarks for coding tasks and autonomous agent capabilities—areas currently dominated by Anthropic and OpenAI.

Sundar Pichai acknowledges Google is trailing in AI coding capabilities

Responding to JPMorgan analyst Doug Anmuth, Pichai stated that Google still holds an advantage in certain respects. He noted, “There are many attributes where we remain at the forefront.” However, Google’s top executive acknowledged that the company had fallen behind in some key areas. “There are areas where we’ve admitted we need to improve, and coding—as well as coding agents—is one example of that,” he added.

To put this in context, coding has rapidly established itself as one of the most important capabilities for AI tools. Companies worldwide are increasingly using tools like Claude Code and Codex to write code. Even Meta and Elon Musk’s SpaceX have joined the AI coding race with new models.

Rather than focusing on the Gemini 3.5 Pro delay during the earnings call, Pichai repeatedly highlighted Gemini Flash—Google’s fastest and most cost-effective model—which is used in cybersecurity, customer service agents, data analysis, and enterprise software. He noted that Gemini 3.6 Flash, launched earlier this week, had improved by more than 10 points on a coding benchmark test compared to the previous version, while also using fewer tokens.

The company also introduced Gemini 3.5 Flash-Lite and a cybersecurity-focused Flash Cyber model, while keeping Gemini 3.5 Pro in the partner testing phase. Their defense reflected a broader effort to convince investors to look beyond a single delayed flagship model and consider Google’s AI ecosystem as a whole—including cloud infrastructure, custom chips, and Gemini products integrated across its various services.

Gemini surpasses 950 million users

This does not mean Google is failing to see a return on its investments. Sundar Pichai stated, “Our AI investments are redefining what is possible across all areas of our business,” adding that the company was seeing “promising momentum across the board.” According to company data, Gemini reached 950 million monthly active users, up from the 750 million recorded in the fourth quarter of 2025.