Google Launches Gemini 3.6 Flash: Faster, Cheaper, and Built for Production AI Agents

Google has released Gemini 3.6 Flash and 3.5 Flash-Lite to general availability, delivering improved token efficiency, stronger coding and agentic performance, and lower prices than previous generations. The release includes a specialized cybersecurity model, 3.5 Flash Cyber, and a teaser that Gemini 4 pre-training has begun.

Key Takeaways

Google has released Gemini 3.6 Flash and 3.5 Flash-Lite to general availability, delivering improved token efficiency, stronger coding and agentic performance, and lower prices than previous generations.
The release includes a specialized cybersecurity model, 3.5 Flash Cyber, and a teaser that Gemini 4 pre-training has begun.

AI synthesis4 sources4

Google has released two new AI models to general availability — Gemini 3.6 Flash and Gemini 3.5 Flash-Lite — bringing faster performance, lower costs, and significant improvements in coding and agentic capabilities to developers and enterprises worldwide.

The announcement, made on July 21, positions 3.6 Flash as the new workhorse model in Google's lineup. According to Google, it reduces output token usage by 17% compared to 3.5 Flash while delivering better results across coding, knowledge work, and multimodal tasks. On the DeepSWE benchmark, 3.6 Flash achieved 49% versus 37% for its predecessor, and on the MLE Bench for machine learning research, it scored 63.9% compared to 49.7%.

"3.6 Flash not only delivers a step up in coding and knowledge work, but it does this while meaningfully improving token efficiency," Google wrote in its announcement. The model is priced at $1.50 per million input tokens and $7.50 per million output tokens — both lower than 3.5 Flash.

Alongside 3.6 Flash, Google launched Gemini 3.5 Flash-Lite, described as the fastest model in the 3.5 series. Running at 350 output tokens per second and priced at just $0.30 per million input tokens, it is designed for high-volume, low-latency agentic workflows. Despite its "Lite" designation, it outperforms the earlier Gemini 3 Flash on several benchmarks including SWE-Bench Pro (54.2% vs. 49.6%) and OSWorld-Verified (74.0% vs. 65.1%).

Both models include built-in computer use capabilities and are available through Google AI Studio, the Gemini API, Gemini Enterprise, and consumer apps. Gemini 3.6 Flash is also rolling out in GitHub Copilot.

In a notable addition, Google introduced Gemini 3.5 Flash Cyber, a specialized model fine-tuned for finding and fixing cybersecurity vulnerabilities. Due to its dual-use nature, it will be available only to governments and trusted partners through a limited-access pilot program.

Google also confirmed that Gemini 3.5 Pro is currently testing with partners and will be made broadly available when ready. Perhaps most significantly, the company revealed that it has begun "the most ambitious pre-training run yet" for Gemini 4.