Builders and clients constructing manufacturing AI brokers want greater token effectivity, decrease latency, and extra dependable efficiency. Our Flash collection of fashions is constructed to satisfy the candy spot of effectivity and high quality to allow scaling agentic workflows. Constructing on Gemini 3.5 Flash, we’re introducing new Gemini fashions:
- 3.6 Flash: Our workhorse mannequin that delivers higher coding, data work, and multimodal efficiency. In response to the Synthetic Evaluation Index, it reduces output token utilization by 17% in comparison with 3.5 Flash, and in some benchmarks like DeepSWE by Datacurve, we observe as much as 65%, all at a decrease value per output token.
- 3.5 Flash-Lite: Our quickest, most cost-effective 3.5-class mannequin, delivering 350 output tokens per second in accordance with the Synthetic Evaluation Index, additionally considerably outperforming prior Flash-Lite generations in agentic workflows.
- 3.5 Flash Cyber in CodeMender: Profitable cybersecurity purposes require cautious orchestration of a mannequin alongside an agent infrastructure. We’re introducing a mixture of a brand new, extremely environment friendly, specialised cyber-focused mannequin paired with our CodeMender code safety agent that delivers aggressive efficiency on the frontier.
Past at present’s releases, Gemini 3.5 Professional is at present testing with companions and we plan to make it broadly obtainable as quickly because it’s prepared. In parallel, our staff is already specializing in constructing the following technology of fashions. We’ve began our most formidable pre-training run but, for Gemini 4, and are excited by the progress.
3.6 Flash: Extra environment friendly and higher high quality than 3.5 Flash
Gemini 3.6 Flash builds straight on developer and buyer suggestions from 3.5 Flash. 3.6 Flash not solely delivers a step up in coding and data work, however it does this whereas meaningfully bettering token effectivity. For instance, on the Synthetic Evaluation Index, we see 3.6 Flash consuming 17% fewer output tokens than 3.5 Flash. It additionally takes fewer reasoning steps and gear calls to perform multi-step workflows.
This enhanced effectivity can be mixed with a lower cost than 3.5 Flash. At $1.50/1M enter tokens and $7.50/1M output tokens, 3.6 Flash reduces the general value per agentic job, making brokers less expensive to construct and run.
