remontandroid
Gemini 3.6 Flash
Gemini 3.6 Flash builds on the release from I/O 2026 and brings improvements based on feedback from developers and customers. The model is 'token-efficient across tasks' – compared to Gemini 3.5 Flash, it consumes 17% fewer output tokens and requires fewer reasoning steps and tool calls to complete multi-step workflows.
The price has been reduced to $1.50 per million input tokens and $7.50 per million output tokens (previously $9). In coding, the model achieves higher accuracy with fewer unwanted code modifications and reduced execution loops. Specifically, on the DeepSWE benchmark, it achieved 49% (up from 37%) and on MLE Bench 63.9% (up from 49.7%). In knowledge work, it scored 1421 on GDPval-AA (up from 1349) and in computer abilities 83% on OSWorld-Verified (up from 78.4%). The knowledge cutoff date was moved from January 2025 to March 2026.
Gemini 3.5 Flash-Lite and Flash Cyber
Gemini 3.5 Flash-Lite is designed for high-throughput, low-latency tasks such as agentic search and document processing. Compared to 3.1 Flash-Lite, it offers significantly better quality at a price of $0.30 per million input tokens and $2.50 per million output tokens. On the Terminal-Bench 2.1 benchmark, it achieved 54% (up from 31%), on GDM-MRCR v2 72.2% (up from 60.1%), and on GDPval-AA v2 1140 (up from 642). It also outperforms model 3 Flash: SWE-Bench Pro 54.2% vs. 49.6%, OSWorld-Verified 74.0% vs. 65.1%.
The company also introduced Gemini 3.5 Flash Cyber, focused on detecting and fixing security vulnerabilities. The model uses multiple Flash Cyber agents in the CodeMender tool. Access is initially for governments and trusted partners as part of a pilot program.
Availability and Outlook
Gemini 3.6 Flash and 3.5 Flash-Lite are available today in the Gemini app, with the Lite version also coming to Search. Developers can use them via Google Antigravity, AI Studio, and Android Studio. Google reiterated that Gemini 3.5 Pro is being tested with partners and will be widely available 'as soon as it is ready'. The DeepMind team has already begun pre-training for Gemini 4 – 'the most ambitious pre-training run to date'.