Gemini 3.8 Flash launch reported

Google introduced Gemini 3.8 Flash on Wednesday, confirming a launch that had circulated earlier in an X post from user Ngxuhby with attached benchmark tables. Google’s official launch post described Gemini 3.8 as its “best reasoning & coding model yet” and said the release includes two variants: Gemini 3.8 Flash and Gemini 3.8 Flash Cyber. The company said the new Flash model keeps the same introductory pricing as Gemini 3.7 Flash at $0.75 per million input tokens and $3.75 per million output tokens. The X post’s broad claim — that Gemini 3.8 Flash scores strongly on reasoning, coding and science benchmarks — lines up with Google’s own description and model-card materials. Google said the model builds on Gemini 3.7 Flash and delivers performance advances across software engineering and knowledge workflows, while the model card says it was evaluated on coding, scientific reasoning, multimodal tasks and long-context use cases. (blog.google) ### What did Google actually launch? Google said Wednesday’s release includes Gemini 3.8 Flash, which it called its “most intelligent workhorse model,” and Gemini 3.8 Flash Cyber, a security-focused variant. The launch post said the release comes three weeks after Gemini 3.7 Flash and marks Google’s third Flash release in six weeks. (blog.google) The Google DeepMind model card published Wednesday describes Gemini 3.8 Flash as the next iteration in the Gemini 3 family and says it supports customizable effort levels to balance quality, cost and latency. Google’s model-card index also lists Gemini 3.8 Flash as newly updated on September 2, 2026. ### Did the benchmark image on X get the story broadly right? (blog.google) Ngxuhby’s post said Gemini 3.8 Flash was strong on reasoning, coding and science-style tasks but weaker on some agentic work. Google’s launch post supports the first half of that characterization, saying the model posts significant improvements over 3.7 Flash in software engineering, critical multi-step reasoning and specialized domains. (deepmind.google) Google’s own materials do not frame the model as broadly weak on agentic tasks. In fact, the launch post says Gemini 3.8 Flash improves on 3.7 Flash across agentic tasks as well, while the model card says the system was evaluated on computer use and related workflow categories in addition to coding and scientific reasoning. The more precise read is that outside benchmark tables may show mixed relative standings against rivals on some agentic tests, but Google’s official description presents 3.8 Flash as an across-the-board upgrade from 3.7 Flash. (blog.google) ### What numbers matter most for developers? Google said Gemini 3.8 Flash is priced at $0.75 per million input tokens and $3.75 per million output tokens, the same introductory price as Gemini 3.7 Flash. That matters because Google is positioning the model as a higher-performing successor without a price increase at launch. (blog.google) The company’s launch post also said 3.8 Flash keeps the speed and low cost profile of 3.7 Flash. That places the release in the same “workhorse” tier Google has used for models meant for production coding, reasoning and interactive application workloads. ### How does this fit Google’s recent release cadence? (blog.google) Google launched Gemini 3.7 Flash about two weeks ago, after launching Gemini 3.6 Flash in July and Gemini 3 Flash in December 2025, according to Google’s blog and model-card pages. Wednesday’s 3.8 Flash release continues a rapid update cycle for the company’s lower-cost Gemini line. (blog.google) Tulsee Doshi, a senior director of product management at Google, said in the launch post that Gemini 3.8 builds on the momentum of 3.7 Flash from three weeks ago. Google’s next reference points are already public: the September 2 launch post, the Gemini 3.8 Flash model card, and the benchmark tables attached to the official announcement materials. (blog.google 1) (blog.google 2)