Google rolls out Gemini 3.6 Flash with faster coding and cheaper pricing

Google has launched a new generation of its Gemini AI models, with a faster and more cost-effective version leading the charge. The company released three new models—though the long-awaited Gemini 3.5 Pro, originally due in June, remains unavailable.

Gemini 3.6 Flash: Faster coding, lower costs#

Gemini 3.6 Flash replaces the 3.5 Flash version that debuted at Google's May I/O conference. The updated model shows measurable improvements in technical tasks, particularly code generation, where it scores 49 percent on the DeepSWE benchmark compared to 37 percent for its predecessor. Computer use—the ability to interact with software interfaces—also improved modestly, from 78.4 percent to 83 percent on the OSWorld test.

More significantly, the new model achieves these gains while consuming roughly 17 percent fewer tokens, the computational units that determine API costs. Pricing has dropped on output tokens ($7.50 per million, down from $9), making the model more economical for developers running agentic workflows that involve multiple steps and decision-making.

Efficiency becomes central#

Google appears to be responding to industry pressure around token consumption and cost. The company has positioned efficiency improvements as central to this release, suggesting that developer concerns about expensive AI inference have shaped the roadmap. The 3.6 Flash now supports computer use as a standard API feature, enabling automated interaction with user interfaces at scale.

Lean and specialized variants#

Alongside the main release, Google introduced Gemini 3.5 Flash Lite—described as the company's most efficient modern AI model, reaching 350 tokens per second. Though pricing crept up slightly ($0.30 per million input tokens, $2.50 per million output tokens), the model is positioned as ideal for teams scaling agentic systems on tight budgets.

Google also launched Gemini 3.5 Flash Cyber, a specialized variant designed for cybersecurity tasks, though few details were provided about its capabilities or pricing.

The 3.5 Pro delay#

Gemini 3.5 Pro, originally announced for June availability, remains in development. The delay follows a pattern of longer timelines for Google's high-end models, even as the company accelerates releases of efficient variants. The continued absence of the Pro tier model suggests the company may still be refining its most capable offering.