Google today announced the launch of Gemini 3.6 Flash, an update to the model unveiled last May at I/O 2026. The new version incorporates developer and customer feedback, offering improved token efficiency across tasks. Compared to Gemini 3.5 Flash, the new model consumes 17% fewer output tokens according to the Artificial Analysis Index, while taking fewer reasoning steps and tool calls to complete multi-step workflows. Pricing is lower: $1.50 per million input tokens and $7.50 per million output tokens, down from $9 per million output.
Superior coding performance and increased precision
In coding tasks, Gemini 3.6 Flash delivers higher precision with fewer unwanted code edits and reduced execution loops. Benchmarks show a significant leap: on DeepSWE the model achieves 49% versus 37% for its predecessor, while on MLE Bench for ML research it hits 63.9% compared to 49.7%. For knowledge work, the GDPval-AA score rises to 1421 from 1349. Computer use capabilities improve from 78.4% to 83% on OSWorld-Verified. The knowledge cutoff date finally advances from January 2025 to March 2026.
Sponsored Protocol
Gemini 3.5 Flash-Lite for high-throughput, low-latency tasks
Google also unveiled Gemini 3.5 Flash-Lite, designed for tasks requiring high throughput and low latency, such as agentic search and document processing. Google claims it offers significantly better quality than the March-released 3.1 Flash-Lite, with pricing at $0.30 per million input tokens and $2.50 per million output tokens. Improvements are evident in Terminal-Bench 2.1 (54% vs 31%), long-context GDM-MRCR v2 (72.2% vs 60.1%), and real-world task execution on GDPval-AA v2 (1140 vs 642). Additionally, 3.5 Flash-Lite outperforms 3 Flash on SWE-Bench Pro (54.2% vs 49.6%) and OSWorld-Verified (74.0% vs 65.1%).
Sponsored Protocol
New security model: Gemini 3.5 Flash Cyber
The company also introduced Gemini 3.5 Flash Cyber, a specialized variant for detecting and fixing security vulnerabilities. Based on Flash for its performance and efficiency, the model aims to identify, validate, and patch code security issues at scale and at a lower per-token cost than larger models. Google's CodeMender tool uses multiple 3.5 Flash Cyber agents. Access is initially limited to governments and trusted partners as part of a restricted pilot program, giving frontline defenders a head start in finding and fixing critical vulnerabilities before they can be exploited.
Sponsored Protocol
Gemini 3.6 Flash and 3.5 Flash-Lite are available today in the Gemini app, with the latter also coming to Google Search. Developers can access them via Google Antigravity, AI Studio, and Android Studio. Looking ahead, Google reiterated that Gemini 3.5 Pro is currently testing with partners and will be released broadly when ready. Meanwhile, the DeepMind team is already working on the next generation, with pre-training for Gemini 4 underway. For more on the AI landscape, check out the article on how Chinese AI divides the White House. Further context is available on Wikipedia's Gemini page.
Source: https://9to5google.com/2026/07/21/gemini-3-6-flash-launch
