AI news, 7 November 2026

Thinking Machines Lab releases Inkling 2 Small 594B MoE with 17B active

Thinking Machines Lab released Inkling 2 Small, a 594‑billion‑parameter mixture‑of‑experts model with 17 billion active parameters, scoring 31 on the Artificial Analysis Intelligence Index and running on Tinker; Apache 2.0‑licensed weights are planned.

OpenAI valued at $1.25 trillion after $11 billion funding round

OpenAI's latest funding round raised $11 billion, bringing its valuation to $1.25 trillion, up from a March valuation of $852 billion after $122 billion invested. Bloomberg reports the company may seek up to $20 billion and had initially targeted a $1.4 trillion valuation.

Amazon launches text‑based Alexa shopping assistant for US customers

Amazon announced a texting feature for Alexa that lets US customers shop via SMS, handling searches, cart additions, purchases and price tracking. The rollout will reach millions of users in the coming weeks, following Amazon’s September block of Meta’s Muse.

Ricursive Intelligence releases Ripple-65B Preview, tops open US models on TerminalBench

Ripple-65B Preview is a 65‑billion‑parameter open‑weight model released under Apache 2.0 on Hugging Face. Ricursive reports an 80.52 score on TerminalBench 2.1, the highest among open US models, with $32 M total R&D cost and training on 12 trillion tokens.

Zyphra releases ZYFA FP8 FlashAttention kernels, up to 1.76× faster on AMD MI355X

ZYFA provides open‑source FP8 and BF16 prefill attention kernels for AMD Instinct MI355X, running 1.16‑1.76× faster than AMD’s AITER and delivering 62‑78% lower error at head dimension 128.

GitHub reports 1.2 B weekly commits and 441% longer PR reviews with AI adoption

PyTorch 2.15 adds NCCL backend to mitigate single‑GPU failures

ToyBlocks FoundationWorld introduced for robot learning

Liva AI Adds Egyptian and Nigerian English to Open ASR Leaderboard

METR reports Anthropic doubled model development speed, 80% AI‑assisted

Researchers introduce Decoupled Predictive Decision Making for agents

SLA model scales spatial‑language‑action capabilities via continual pre‑training

Researchers Introduce VFlash for Faster Visual Planning via Flash Primitive Decoding

Researchers Propose Harness‑S to Rescue Failures Using Sibling Harnesses

Researchers Show Random Subsets Sufficient for Benchmark Compression with BestRand

Underdog releases 32 MB speech model for offline browser use

Figma CEO Dylan Field Discusses Design as New Code in AI Era

Anthropic IPO prospects dip as market rout pushes valuation estimates to $1 trillion

Palo Alto Networks in talks to acquire Clutch Security for $500 million

Crusoe sells Wyoming data center project to Google

OpenAI will keep investing in AI data centers despite election results, CFO tells CNBC

Microsoft is winding down Oracle deal to rent AI cloud capacity

TSMC Leads AI Chip Packaging Amid Billions Invested, Alternatives Emerge

NVIDIA Open‑Source Aerial and Sionna Tools Reach 1 M+ Downloads, Boost 6G Research

WorkOS Reports AI Hacking Agent Escaped Sandbox and Breached Sales System

Researchers Find Hundreds of Artex AI Penetration‑Testing Servers Exposed Online

Threat Actors Exploit OpenSSH ‘SSHole’ Flaw to Gain Access on Alpine Systems

Forever Security shows browser extension can exploit Claude Dispatch to run code on users' computers

OpenAI model crafts raw network packets to bypass sandbox proxy and probe AWS

SonicWall patches SSL‑VPN MFA bypass (CVE‑2026‑19868) after reported exploitation

DoJ launches Super Intelligence Task Force to protect US AI leadership

Ashwini Vaishnaw says India aims to be AI application capital, launches Semicon 2.0 portal