Google has released its third Gemini Flash model in six weeks, Gemini 3.8 Flash, which the company claims is its best reasoning and coding model yet. The model comes in two variations: the standard Flash, described as a 'workhorse' for tasks like agentic work and software development, and Gemini 3.8 Flash Cyber, tailored for vulnerability detection and mitigation. Google is offering API access at an introductory rate of $0.75 per million input tokens and $3.75 per million output tokens, with regular pricing set at $1.50 and $7.50 respectively. The lower prices are likely a strategic move to retain business interest amid competition from other AI labs. Source: arstechnica
Google provided benchmark results showing Gemini 3.8 Flash competing with or outperforming larger, more expensive models. The model shows marginal improvements over Gemini 3.7 Flash in most tests, with more significant gains in coding evaluations. It now leads the DeepSWE leaderboard for solving complex software engineering problems at a lower cost. However, it still lags behind the market leader, Claude Opus, in the OSWorld-2.0 test of agentic computer use. Despite this, Google claims its new Flash models are now on par with market leaders in certain areas. Source: arstechnica
Gemini 3.8 Flash Cyber, replacing the 3.5 version, has demonstrated substantial improvements in internal testing, identifying more vulnerabilities and issuing working patches more frequently. The Chrome security team reported a 2.6x increase in patch accuracy with the new model, while the Cloud team found a critical vulnerability in just two hours. Partners like Wiz and Palo Alto Networks have also praised the model's capabilities. Gemini 3.8 Flash will be available across Google's ecosystem, while Cyber is currently limited to trusted testers and governments. Source: arstechnica
