AI Daily · Oct 1, 2026: Google DeepMind rolls out Gemini 4 Argon to trusted testers
Google DeepMind announced Gemini 4 Argon, rolling out to trusted testers via the Fairwind Program with an industry-leading 1M token output limit, while Perplexity's contextual embedding model pplx-embed-v2-context-9b-preview set new ConTEB records at 1/8 the storage of voyage-context-4. Other notabl
Showing the latest briefing (Oct 1, 2026). Every day is generated the next morning.
Pick a date
-
Google DeepMind rolls out Gemini 4 Argon to trusted testers
Google DeepMind announced Gemini 4 Argon, a frontier model targeting complex workflows in coding, enterprise knowledge work, and cybersecurity defense, rolling out today to a set of trusted testers through the Fairwind Program. It also reports an industry-leading 1M token output limit.
-
Perplexity releases contextual embedding model pplx-embed-v2-context-9b-preview
Perplexity introduced pplx-embed-v2-context-9b-preview, a contextual embedding model that encodes each chunk with the full document in view. It sets a new state of the art on ConTEB and turbopuffer's privately held context-bench, and on ConTEB achieves the highest average nDCG@10 among tested models, beating voyage-context-4 on chunk retrieval with 1 KB per vector (1024 dims, int8) versus 8 KB (20
-
OpenAI requires hardware keys for cyber capabilities on individual plans
Starting October 1, individual Daybreak Blue users need an eligible hardware key to use cyber-forward capabilities, while enterprise users are unaffected. OpenAI says the keys help verify account holders and support accelerating defenders.
-
StepFun team member's X account hacked for phishing
A StepFun teammate's X account was hijacked, with the handle changed to @ContentCase and used to distribute phishing messages before being suspended; recovery efforts are ongoing.
-
Gemini launches Skills globally, replacing Gems later
Google is rolling out Skills—saved instruction shortcuts invoked via '/' in the prompt bar—globally in the Gemini app today, with Workspace business, enterprise, nonprofit, and education customers getting it in coming weeks. Gems will auto-migrate to Skills, fully replaced for personal accounts starting November (March 2027 for Workspace, June 2027 for Work accounts).
-
Perplexity opens free email-based Computer agent access
Perplexity is opening Computer agent access without requiring an account, and running all tasks delegated via email to computer@perplexity.com for free for a limited time, with results viewable on web and mobile.
-
HeyGen Video One Launches On OpenRouter Platform
HeyGen's Video 1, built on MiniMax H3 and post-trained by HeyGen, is now available on OpenRouter for text-to-video, image-to-video, and reference-to-video generation at $0.02 per rendered second ($0.01 with 50% off through October), producing 5-15 second clips at 480p or 768p.
-
Pareto 26.10 preview launches on OpenRouter with lower pricing
Pareto 26.10 Preview from theunbiasedco is available on OpenRouter at $0.80/M input and $3.20/M output, versus $2.50/$7.50 for the current Pareto, offering 1M context, text and image input, and tool calling for research, coding, and agentic workflows.
-
Voyage AI rerank-3 and rerank-3-lite now available on OpenRouter
OpenRouter added Voyage AI rerank-3 and rerank-3-lite as drop-in upgrades to rerank-2.5, priced at $0.05 and $0.02 per million tokens respectively, each handling 32K tokens per query-document pair.
-
OpenRouter launches Security Center after auditing 1,000 plus API keys
After finding over 1,000 active API keys across 85 employees with many unused for months, OpenRouter built a Security Center on all plans that flags risky keys and lets users disable, archive, or cap up to 500 at once.