Foundation Models - News Articles

Microsoft unveils MAI-Transcribe-2-Streaming for real-time transcription in 60 languages

Oct 05, 2026

The Deep View

Microsoft unveils MAI-Transcribe-2-Streaming, its first real-time transcription model for 60 languages, saying it produces text within 100 milliseconds for live dictation and subtitles, alongside new MAI-Voice-2.1 text-to-speech models covering 23 languages and 26 locales.

Audio & Voice Foundation Models

Strands Launches Open-Source Decider 2B for Low-Latency Local Agent Decisions

Oct 02, 2026

Strands Agents

Strands launches the open-source Decider 2B model for low-latency local AI-agent decisions, reporting roughly 115 ms median latency on an Nvidia RTX 3090 and a top JevBench ranking among strictly 2B-parameter models.

Agents Open Source Foundation Models

arXiv Imposes Two-Paper Monthly Submission Limit as AI-Written Manuscripts Surge

Oct 01, 2026

News from arXiv

arXiv imposes a universal two-paper-per-month submission limit starting October 1, 2026, responding to a surge in AI-written, fragmented manuscripts that helped drive 40,363 submissions and nearly 9,000 support tickets in September.

Research Regulation Foundation Models

Google Unveils Gemini 4 Argon, Rolls Frontier Model Out to Cyber Defenders

Oct 01, 2026

The Deep View

Google unveils Gemini 4 Argon, a frontier model now rolling out to cyber defenders through its restricted Fairwind program, with a 1 million-token output limit and introductory pricing of $2 per million input tokens and $10 per million output tokens.

Foundation Models Security

Google Unveils Gemini 4 Argon, Claims Lead Over Rivals in Coding, Knowledge Work and Cybersecurity

Oct 01, 2026

The Deep View

Google unveils Gemini 4 Argon, a flagship AI model it claims leads rivals in coding, knowledge work and cybersecurity, with restricted Fairwind access for cyber defenders and pricing starting at $2 per million input tokens.

Foundation Models Coding & Developer Tools Security AI Safety

CrowdStrike Says AI Shifts Cybersecurity Balance Toward Defenders as It Helps Disrupt Sality Botnet

Oct 01, 2026

The Deep View

CrowdStrike says AI is beginning to shift cybersecurity’s advantage toward defenders as it and partners disrupt the two-decade-old Sality botnet, using Nvidia Nemotron-based red- and blue-team models trained on CrowdStrike security data.

Security Foundation Models

Google Unveils Gemini 4 Argon, Giving Trusted Cyber Defenders a Model It Says Can Autonomously Patch Critical Flaws

Oct 01, 2026

Google

Google unveils Gemini 4 Argon, a cyber-defense model it says can autonomously find, validate and patch critical vulnerabilities, initially restricting access to trusted defenders in its Fairwind Program after Wiz uses it to uncover a healthcare-software flaw earlier frontier models missed.

Security Foundation Models Coding & Developer Tools

Google Unveils Gemini 4 Argon for Select Cybersecurity Teams, Claiming Benchmark Lead Over GPT-6 Astra and Claude Opus 5.5

Oct 01, 2026

The Rundown AI

Google unveils Gemini 4 Argon for vetted cybersecurity teams, claiming it leads GPT-6 Astra and Claude Opus 5.5 on 13 of 19 internal benchmarks and reaches 77.9% on DeepSWE coding.

Foundation Models Security Coding & Developer Tools

Fig launches RIDGE, finding frontier AI models have jagged task performance with no clear overall leader

Sep 30, 2026

Fig.inc

Fig launches RIDGE, finding no clear frontier-AI champion as Astra, Opus 5.5 and rivals show jagged task-level results; Astra leads VisualWebArena at 139 of 177 tasks, yet every lower-scoring model solves some tasks it misses.

Foundation Models Research Agents

PageIndex SDK Adds Local Indexing, Retrieval and Chat With Users’ Own LLM Keys

Sep 30, 2026

GitHub

PageIndex SDK now lets developers index, retrieve from, and chat with text-based PDFs locally using their own LLM keys while retaining access to PageIndex Cloud through the same client, with PageIndex Flash becoming its default local tree-indexing method.

Coding & Developer Tools Data & Analytics Foundation Models