Google DeepMind rolled out three new Gemini Flash models on Tuesday, all built to make AI agents cheaper and faster, while its delayed flagship stayed in partner testing.
Key Points:
- Google shipped Gemini 3.6 Flash, 3.5 Flash-Lite and a security-tuned 3.5 Flash Cyber on Tuesday.
- The workhorse 3.6 Flash uses up to 17% fewer output tokens than its predecessor.
- The long-promised Gemini 3.5 Pro stayed absent, still in testing with partners.
Gemini Flash Lineup Expands
The company confirmed the trio in a Tuesday post, casting 3.6 Flash as its main workhorse for coding, knowledge work and multimodal tasks.
All three sit in the cheaper Flash tier, tuned for speed and high volume rather than the deep reasoning of a top model. Google said the aim was efficiency, latency and reliability for customers building AI agents at scale.
The flagship Flash model uses up to 17% fewer output tokens than the version it replaces, which trims the cost of running it. Google now prices it at $1.50 per million input tokens and $7.50 per million output, cheaper than the $9 the previous model charged, and moved its knowledge cutoff to March 2026.
The cheaper Flash-Lite targets high-volume jobs such as document processing and agentic search, and slots in below its sibling on price. The third model was the outlier. Flash Cyber hunts and patches software flaws inside Google's CodeMender agent, and the company found it flagged 55 confirmed vulnerabilities in a test of the V8 JavaScript engine, though access stays limited to governments and trusted partners.
Also Read: Anthropic Chased An $11B Robot Startup Before The Denial Went Public
Gemini 3.5 Pro Delay Weighs
What Google left out drew as much notice as what it shipped. The flagship 3.5 Pro, promised for June at the company's May developer conference, stayed in partner testing with no firm date, months behind the timeline Google set on stage two months earlier.
Product lead Logan Kilpatrick said the team hoped to land Pro soon and had already begun its most ambitious pre-training run yet for Gemini 4.
Rivals are not waiting, though. In barely a week, OpenAI, Anthropic and others pushed out fresh models, and Alphabet shares slipped almost 6% from Thursday through midday Tuesday.
The pattern is familiar for Google, which has leaned on cheaper, faster Flash models to stay in the mix while its top-tier release keeps slipping. Back in May, the company billed 3.5 Pro as nearly ready, yet the flagship gap has only widened as faster-moving rivals keep shipping.
Read Next: Hulu's Charles Manson Documentary Uncovers His Unknown Granddaughter As Prediction Markets Watch