Google expands Gemini lineup with three new models and broader Spark access

Three new Gemini models released

Google announced three new artificial intelligence models — Gemini 3.6 Flash, Gemini 3.5 Flash-Lite and Gemini 3.5 Flash Cyber — targeting coding, AI agent workloads and cybersecurity analysis respectively. Gemini 3.6 Flash builds on Gemini 3.5 Flash with improved coding performance, stronger research capabilities and faster multi-step reasoning while reducing token usage by 17%, according to Google.

Pricing and benchmark claims

Gemini 3.6 Flash is priced at US$1.50 per million input tokens and US$7.50 per million output tokens. Google said benchmark tests show the model outperforms Gemini 3.5 Flash in coding, AI agent tasks, computer control and knowledge-based research, offering a better balance of speed, accuracy and efficiency for complex workloads. The lighter Gemini 3.5 Flash-Lite is aimed at search, large-scale data processing and AI agent applications, can generate up to 350 output tokens per second and is priced at US$0.30 per million input tokens and US$2.50 per million output tokens.

Cybersecurity model and access limits

Gemini 3.5 Flash Cyber is based on an optimised version of the Gemini 3.5 Flash architecture to perform security analysis faster while reducing token usage per request. According to Google, the model will initially be available exclusively through CodeMender, with access limited to government agencies and trusted organisations. Google also said an update on its flagship Gemini 3.5 Pro model will be announced in the near future.

Gemini Spark reaches AI Pro subscribers

Google began rolling out Gemini Spark, its AI assistant that can perform tasks and monitor activities in the background, to Google AI Pro subscribers in the United States, where the plan costs $20 per month. The feature had previously been available mainly through Google's higher-priced AI Ultra subscription. Announcing the update on X, Google said it would bring Gemini Spark to AI Pro subscribers in more countries soon. New capabilities include editing private and shared Google Docs, Sheets and Slides, reading comments in spreadsheets and presentations, and adding images via the Canvas tool. Earlier in July, Google extended Spark support to AI Ultra subscribers in all regions where Gemini Apps are available except the European Economic Area, Nigeria, Switzerland and the United Kingdom.

Gemini 3.5 Pro delay and user growth

Reports indicate Alphabet is still testing its Gemini 3.5 Pro model, announced in May, and has not considered it ready for general release. Monthly Gemini users reached 900 million in May, more than doubling in less than a year, and the Gemini product generated roughly $1.2 billion in sales last year. Apple is reported to be paying Alphabet about $1 billion annually to use Gemini as the underlying AI model for the new Siri AI. Separately, Alphabet is designing a chip called Frozen v2 that reportedly incorporates parts of the Gemini architecture directly onto the processor, delivering 6 to 10 times more AI tokens per unit of power than its current Tensor Processing Units, with a launch expected in 2028.