Google Releases New Gemini Flash Models for Agents

Google released three new Gemini models on May 21, 2024, designed to enhance performance for AI agents by focusing on token efficiency, lower latency, and cost-effectiveness. The new lineup includes Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, and Gemini 3.5 Flash Cyber, all part of Google's Flash tier, which prioritizes speed and high-volume agentic tasks over maximum reasoning depth.

Gemini 3.6 Flash, positioned as the new default model, offers improved quality and efficiency compared to its predecessor, Gemini 3.5 Flash. Google reports that Gemini 3.6 Flash uses 17% fewer output tokens on the Artificial Analysis Index and up to a 65% reduction on the DeepSWE benchmark. It also requires fewer reasoning steps and tool calls in multi-step workflows. This increased efficiency translates to lower costs, with Gemini 3.6 Flash priced at $1.50 per 1 million input tokens and $7.50 per 1 million output tokens, a reduction from the previous $9.00 output rate for 3.5 Flash.

Beyond efficiency, Gemini 3.6 Flash demonstrates significant quality improvements. On the DeepSWE benchmark, it achieved a score of 49%, up from 37% for 3.5 Flash. Other benchmark scores include 63.9% on MLE Bench (compared to 49.7%), 83.0% on OSWorld-Verified (compared to 78.4%), and 1421 on the GDPval-AA v2 knowledge-work benchmark (compared to 1349).

Google has also integrated computer use as a built-in client-side tool through the Gemini API and Gemini Enterprise. Early adopters like Hebbia and Harvey have reported benefits in areas such as document parsing, chart and data analysis, and report drafting. Gemini 3.6 Flash is also shipped with enhanced Frontier Safety safeguards, specifically addressing Chemical, Biological, Radiological, and Nuclear (CBRN) threats and cyber-offense misuse.