Google Begins September with AI Momentum After Longest Monthly Losing Streak in Over a Decade After concluding its longest monthly losing streak on Wall Street in more than a decade, Google has entered September with a series of developments aimed at rekindling investor confidence. The company launched Gemini 3.8 Flash, its third Flash model in six weeks, alongside a new cybersecurity model tailored for trusted government and enterprise clients. These moves, combined with a federal judge’s rejection of the Justice Department’s attempt to force Google to divest its ad exchange, mark a turning point for Alphabet amid a challenging summer marked by AI competition, talent departures, and internal restructuring at DeepMind. The stock’s recovery has been modest, with a 0.6% rise on Wednesday following a 1% drop on Tuesday, leaving it slightly down for the month. However, Google is emphasizing its advancements in Gemini 3.8 Flash, which focuses on coding and agentic tasks—areas where AI companies are striving to convert technological progress into enterprise revenue. The model is described as Gemini’s best reasoning and coding iteration to date, with notable improvements over its predecessor, Gemini 3.7 Flash, particularly in software engineering and multi-step tasks. Tulsee Doshi, senior director of product management at Google DeepMind, told CNBC that recent Flash models have “really surprised us in positive ways in their performance,” offering opportunities to leverage these advancements. While smaller Flash models are cheaper and faster to iterate than Google’s largest frontier systems, analysts like Gil Luria of D.A. Davidson note that they still fall short of competing with Anthropic or OpenAI.#google #alphabet #google_cloud #fairwind_program #deepmind
Google Introduces Gemini 3.8 Flash and Cyber Models for Enhanced AI Capabilities Google DeepMind has launched two new iterations of its Gemini series, Gemini 3.8 Flash and Gemini 3.8 Flash Cyber, designed to advance agentic workflows and cybersecurity applications. These models build on the previous 3.7 Flash release, which debuted three weeks prior, and represent the third Flash update in six weeks. Both variants are powered by the same foundational intelligence but are optimized for distinct use cases, with the Cyber variant tailored for security-focused environments. Gemini 3.8 Flash is positioned as an improvement over its predecessor, offering enhanced reasoning and coding capabilities without compromising speed or cost. The model demonstrates significant performance gains in complex engineering tasks, outperforming many higher-cost frontier models on the DeepSWE v1.1 benchmark. This benchmark evaluates the model’s ability to autonomously solve intricate software engineering problems, with Gemini 3.8 Flash achieving results comparable to more expensive alternatives. Additionally, it excels in specialized domains such as finance and law, surpassing 3.7 Flash and other models in benchmarks like Vals Finance Agent V2 and Harvey’s Legal Agent Benchmark. The model’s performance on the HLE-Verified benchmark, where it achieved a 54.9% score, highlights its capability to handle multi-step reasoning across STEM, humanities, and professional fields. The improvements in Gemini 3.8 Flash stem from its design to execute more thorough reasoning steps, particularly on complex tasks. The model may use additional tokens at higher effort levels to maximize performance, though developers can opt for lower effort settings to reduce computational overhead.#google #google_deepmind #gemini_3_8_flash #gemini_3_8_flash_cyber #fairwind_program
