Google announces Gemini 3.6 Flash and cybersecurity AI, teases 3.5 Pro and Gemini 4 - Ars Technica

What the report says

Ars Technica reported that Google has introduced three Gemini model updates, led by Gemini 3.6 Flash, while confirming that the long-awaited Gemini 3.5 Pro is still not publicly available. The announcement follows Google’s I/O presentation in May, where Gemini 3.5 Flash had been highlighted, but that model is now being replaced in the Gemini app and API by 3.6 Flash.

According to the report, Google says Gemini 3.6 Flash is more capable, improves coding performance, and supports multimodal and computer-use features through the Gemini API. Ars cited Google benchmark figures showing DeepSWE coding performance rising to 49 percent from 37 percent for 3.5 Flash, while an OSWorld computer-use score increased to 83 percent from 78.4 percent. Google also says the model uses about 17 percent fewer tokens, with API output pricing reduced to $7.50 per million tokens while input pricing remains $1.50 per million.

Google also released Gemini 3.5 Flash Lite, described as a lower-cost, high-speed model for scaled agentic systems, with claimed throughput of 350 tokens per second. Ars reported that it will be available to developers and in the Gemini app, and that Google expects it to be used heavily in Search, including AI Overviews.

The third model, Gemini 3.5 Flash Cyber, is Google’s first Gemini version tuned for cybersecurity. Ars reported that Google plans a limited rollout through DeepMind’s CodeMender agent for trusted partners and governments, citing concerns about dual-use risks. Google also said Gemini 3.5 Pro remains in partner testing and that pre-training has begun for Gemini 4, though no release timetable was given.