Google Releases Gemini 3.5 Flash Model: Output Speed 4x Faster Than GPT-5.5

At the 2026 Google I/O Developer Conference, Google CEO Sundar Pichai announced the launch of the Gemini 3.5 Flash model, which outperforms Gemini 3.1 Pro in multiple benchmark tests.

The model's output speed has improved significantly—it is 4x faster than Claude Opus 4.7 and GPT-5.5 xhigh models, reaching 289 tokens per second.

Google's internal team also tested with the Antigravity tool, building a runnable operating system core from scratch in just 12 hours, launching 93 sub-agents and generating 2.6 billion tokens during the process.