Google Releases Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, and Gemini 3.5 Flash Cyber

Jul 21, 2026

Google has released three lightweight AI models—Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, and the cybersecurity-focused Gemini 3.5 Flash Cyber—to offer cheaper, faster alternatives for enterprise workloads. However, Google's highly anticipated flagship model, Gemini 3.5 Pro, remains delayed past its original June 2026 launch target due to internal software coding performance challenges. While competitors like OpenAI and Anthropic ship advanced models, Google is competing on cost efficiency and has commenced pre-training for Gemini 4.

Gemini 3.6 Flash release

▪Gemini 3.6 Flash carries a knowledge cutoff of March 2026, an advance from the January 2025 cutoff of Gemini 3.5 Flash.
▪Google released Gemini 3.6 Flash on July 21, 2026, as a direct successor to the Gemini 3.5 Flash model.
▪Gemini 3.6 Flash is priced at $1.50 per million input tokens and $7.50 per million output tokens, which is cheaper on output than the $9 charged by Gemini 3.5 Flash.

Gemini 3.5 Pro delay

▪Google confirmed that Gemini 3.5 Pro is currently undergoing testing with partners and the United States government.
▪Google delayed the public release of its flagship Gemini 3.5 Pro model, which was originally scheduled for a June 2026 launch.
▪The delay of Gemini 3.5 Pro is reportedly due to the model falling short of internal performance goals in software coding capabilities and hallucination rates.

Gemini 3.5 Flash-Lite launch

▪Google launched Gemini 3.5 Flash-Lite on July 21, 2026, as its fastest and most cost-effective 3.5-series model for high-throughput workloads.
▪Gemini 3.5 Flash-Lite is priced at $0.30 per million input tokens and $2.50 per million output tokens.

Flash Cyber cybersecurity model

▪Google released Gemini 3.5 Flash Cyber on July 21, 2026, a specialized model fine-tuned to scan, verify, and patch software vulnerabilities, running within the CodeMender agent.
▪Gemini 3.5 Flash Cyber is restricted to governments and trusted partners under a limited-access pilot and runs inside Google DeepMind's CodeMender tool.

Benchmark performance comparisons

▪Gemini 3.6 Flash scores 83% on the OSWorld-Verified computer-use benchmark, compared to 78.4% for Gemini 3.5 Flash, and 1421 vs. 1349 on GDPval-AA (knowledge work).
▪Gemini 3.6 Flash scores 49% on the DeepSWE software engineering benchmark, compared to 37% for Gemini 3.5 Flash.
▪Gemini 3.5 Flash-Lite scores 1140 on the GDPval-AA v2 benchmark, compared to 642 for the previous Gemini 3.1 Flash-Lite model, and also outperforms Gemini 3 Flash on SWE-Bench Pro (54.2% vs. 49.6%) and OSWorld-Verified (74.0% vs. 65.1%).

Pricing competitive positioning

▪Google has begun its pre-training run for its next-generation flagship model, Gemini 4.
▪Google CEO Sundar Pichai stated that companies could save upwards of $1 billion per year by shifting most of their AI workloads to Gemini models.