Google has launched beta 'Live' voice features across Gmail, Docs and Keep, letting subscribers talk to these apps instead of typing. Users can ask Gmail to find or summarize emails, have Keep organize notes from spoken input, and tell Docs to draft or pull in content from Gmail, Chat and Drive. The features, first shown at Google's I/O conference in May, are currently limited to AI Plus, Pro and Ultra subscribers, with Docs support restricted to Pro and Ultra tiers.
cnet.com
· 2026-09-04
Google has launched Gemini 3.8 Flash, its third Flash-tier model release in six weeks, which it says leads on the DeepSWE coding benchmark. A specialized version, Gemini 3.8 Flash Cyber, is tuned for vulnerability detection but is currently restricted to trusted testers and government users. The standard model rolls out across Google's ecosystem today at discounted API pricing through year-end.
tech.slashdot.org
· 2026-09-03
Google unveiled Gemini 3.8 Flash, its third Flash-tier model release in six weeks, alongside a new cybersecurity model for government and enterprise clients. The launch coincided with a federal judge rejecting the Justice Department's demand that Google divest its ad exchange, and with Berkshire Hathaway's Greg Abel publicly praising Alphabet's AI standing.
cnbc.com
· 2026-09-03
Google released two new AI models on Wednesday: a standard Gemini 3.8 Flash built for agentic tasks, coding, and multi-step reasoning, and Flash Cyber, tuned specifically for finding and fixing software vulnerabilities. CEO Sundar Pichai said 3.8 Flash beats many larger frontier models on the DeepSWE coding benchmark at lower cost, while Flash Cyber scored 86.2% on the CyberGym benchmark and found vulnerabilities across 20 programming languages with over 70% success in internal testing.
venturebeat.com
· 2026-09-03
Google has launched Gemini 3.8 Flash, its third Flash-tier model release in just six weeks, alongside a specialized variant called Gemini 3.8 Flash Cyber aimed at vulnerability detection and mitigation. The company touts it as its strongest reasoning and coding model to date, with introductory API pricing of $0.75 per million input tokens and $3.75 per million output tokens through year-end before doubling.
arstechnica.com
· 2026-09-02
Google has launched Gemini 3.8 Flash, a successor to Gemini 3.7 Flash released just weeks earlier, which the company says performs more reasoning steps and iteratively calls tools on complex tasks. Input and output token pricing remains unchanged from the prior model, but Google cautions that the new model may consume more tokens overall to boost performance, particularly at higher effort settings. Early analysts note it beats rivals like Anthropic's models on coding, finance, and legal benchmarks, though it can end up pricier in practice due to higher token usage.
theverge.com
· 2026-09-02
Google launched Gemini 3.8 Flash, its third Flash-tier model in six weeks, improving on coding, agentic tasks and multi-step reasoning while keeping the same pricing as 3.7 Flash ($0.75/$3.75 per million input/output tokens). Alongside it, Google introduced Gemini 3.8 Flash Cyber, a specialized version tuned for vulnerability detection and automated patching, accessible only to vetted security teams through a new invite-only Fairwind Program.
blog.google
· 2026-09-02
Google published documentation outlining the intended uses and known limitations of Gemini 3.8 Flash, a model aimed at cost-effective, production-scale agent deployment for developers and enterprises. The company flagged persistent issues including hallucinations, occasional slowness or timeouts, and inconsistent knowledge freshness across domains despite a stated March 2026 cutoff.
deepmind.google
· 2026-09-02
Google has upgraded Gemini's video analysis with a new 'agentic' approach that lets the AI selectively choose which frames, audio, or transcript segments to review instead of processing every frame. The company says this cuts token usage by up to 88% while improving accuracy by up to 7%, and enables tracking of object counts, movement, and sub-second events across multi-hour videos. The feature currently works through the Gemini API in AI Studio and the Enterprise Agent Platform, with plans to reach the Gemini app and YouTube's 'Ask YouTube' tool soon.
androidauthority.com
· 2026-09-02
Meta unveiled Muse Voice Transcribe, its first real-time audio perception model, capable of transcribing over 20 speakers simultaneously while switching between more than 20 validated languages, including mid-sentence code-switching. CEO Mark Zuckerberg demoed the tool on X, noting it uses adaptive delay to balance speed and accuracy. The model is now available through Meta's Mac AI app, Muse Code, and its Model API at $3 per 1,000 audio minutes.
engadget.com
· 2026-09-01
Internal testing of Google's upcoming Gemini 3.8 Flash model reportedly shows meaningful improvement in coding performance, an area where the company has trailed rivals Anthropic and OpenAI. The model is expected to be released publicly this week.
wsj.com
· 2026-09-01
A new study from Google Research and Technion shows that when models like GPT-5 and Gemini-3 hallucinate, the underlying fact is usually already stored in their parameters rather than missing entirely. The researchers found these models encode 95-98% of tested facts, but often fail to surface them during generation due to retrieval, not storage, limits. They propose 'fact-level profiling' to separate whether a fact is encoded versus reliably retrievable across different phrasings.
venturebeat.com
· 2026-09-01
GoKawiil