Google to Unveil New Gemini Model at I/O, Performance Comparable to GPT-5.5

According to tech journalist Alex Heath, citing insider sources, Google plans to announce a new Gemini model at its upcoming annual I/O developer conference. The model is expected to debut this Tuesday, with performance levels roughly comparable to OpenAI's recently released GPT-5.5, though it does not aim to push the frontier of capabilities. This news has drawn widespread industry attention, seen as Google's latest move in the intense large model competition.

Sources indicate that while the new Gemini model performs solidly in benchmarks, it still lags noticeably behind Anthropic's yet-to-be-widely-released Mythos model. Although Mythos has not been broadly commercialized, it has already reshaped how labs define "leading," resetting industry expectations for top-tier model capabilities. At this point, specific details on parameter count, application scenarios, and pricing for the new Gemini model have not been disclosed.

This release comes at a critical time as Google seeks to solidify its position in AI assistants and cloud services. Previous reports have suggested that Apple's next-generation Siri might leverage Google's Gemini model for feature upgrades. Additionally, Google has recently been planning to adjust its Gemini subscription plans to fill pricing gaps and allow users to track AI usage. If this new model successfully addresses the market's demand for mid-to-high-end AI capabilities, it could unlock more business opportunities for Google. However, facing dual pressure from GPT-5.5 and Anthropic's Mythos, it remains to be seen whether Google can regain its technological edge with this release.