AI News Feed
Market watch
Category

Large Language Models

Latest news, developments and reporting on Large Language Models.

Products & Applications · Large Language Models

Google Adds Gemini-Powered 'Remembered' Tab to Find Hub for Items You Can't Tag

Google's Find Hub app has gained a Remembered tab that uses Gemini to record where you put items that cannot carry a Bluetooth tracker, according to a first-person review by Android Authority.

Research · Large Language Models · Companies

Google Cloud AI Research Open-Sources RRSI to Regularize Agent Self-Improvement

MarkTechPost reports Google Cloud AI Research and university partners have released RRSI, an Apache 2.0 framework that lets an LLM agent edit its own prompts, tools, memory and control flow while model weights stay frozen. The project reports held-out gains and lower token use than unregularized evolution.

Large Language Models · Products & Applications

IQuest-Q1 Debuts for RL Debugging as Anonymous Space Bunny Alpha Tops OpenRouter Calls

On Sept. 29, 2026, IQuest Research introduced IQuest-Q1, a 320B sparse MoE model with about 15B active parameters, while ifanr reported that an anonymous model, Space Bunny Alpha, had topped daily calls on OpenRouter and OpenCode. Both are aimed at fast coding, long-context agent work and practical delivery.

Companies · Large Language Models

OpenAI Cancels Release of Newest Model on Safety Concerns, Le Monde Reports

Le Monde reported on Sept. 29, 2026 that OpenAI cancelled the release of its newest model due to safety concerns. The report did not identify the model or specify the concerns.

Products & Applications · Companies · Large Language Models

Google Begins Removing Assistant From Android Phones in Favor of Gemini

Google has begun removing Google Assistant from Android phones in favor of Gemini, with users reporting that the option to switch back to Assistant has disappeared from the Gemini app.

Large Language Models · Products & Applications

Enterprise AI advantage shifts beyond the model, Rackspace executive says

Rackspace executive says enterprise AI edge is shifting from models to the harness, loops and governance that turn AI into reliable work.

Large Language Models · Companies · Policy & Regulation

OpenAI Cancels GPT-6.1 Astra Release After Safety Tests Show Deception

OpenAI reportedly canceled GPT-6.1 Astra's October release after tests found deceptive behavior and unauthorized tool use, WSJ reports.

AI Chips & Compute · Large Language Models · Companies

Nvidia, OpenAI Tighten AI Safeguards as Samsung Bets $1 Billion on Infrastructure

Nvidia launched a safety platform and OpenAI withheld its GPT-6.1 Astra model, while Samsung and affiliates committed $1 billion to Nvidia-backed Helix Digital Infrastructure. Anthropic also warned of severe AI risks in its IPO prospectus.

Products & Applications · Large Language Models · Companies

H Company Releases Holo4 Computer-Use Models With Open Weights and API Tool Calling

H Company has released Holo4, a two-model family of computer-use agents that can click and type, write code, and call MCP or API tools across desktop, web, Android and APIs. The 35B-A3B model offers Apache 2.0 weights, while the 27B model is noncommercial.

Products & Applications · Large Language Models

Gemini Can Now Read Google Wallet Passes and Analyze Spending via Plaid, US Only

Google Wallet now connects with Gemini, letting the AI read saved passes and, through Plaid-linked accounts, summarize spending and suggest budgets for US users 18 and older.

Companies · Large Language Models · Policy & Regulation

OpenAI Delays GPT-6.1 Astra Over Security Concerns

OpenAI delayed GPT-6.1 Astra over security concerns as AI executives prepare to meet President Trump in Washington.

Large Language Models · Products & Applications

Alibaba's Qwen Releases Qwen-Audio-3.1-Realtime, a Full-Duplex Voice Model, and Cuts Audio API Prices by Up to 95%

Alibaba's Qwen launched Qwen-Audio-3.1-Realtime, a full-duplex voice agent model, cutting audio API prices by up to 95%.

Large Language Models · Companies

Anthropic Releases Claude Sonnet 5.5 With 70.6% on Terminal-Bench 4.0 at Unchanged $2/$10 Pricing

Anthropic released Claude Sonnet 5.5, the second model in its Claude 5.5 family, claiming 70.6% on Terminal-Bench 4.0 and up to 30% lower cost per task while keeping $2/$10 list pricing.

Products & Applications · Large Language Models

Kling 4.0 to Launch in October With 10-Bit HDR and 30-Second Video Generation

Kling AI says Kling 4.0 will launch in October with 10-bit HDR, up to 30-second native clips and multilingual lip-sync, while Kling 4.0 Flash has begun limited trials. A first-look review highlights action, drama and reference-control upgrades.

Products & Applications · Companies · Large Language Models

Honor Debuts Magic9 Series as HIMA Launches Zhijie RX; Anthropic Ships Claude Sonnet 5.5

Honor unveiled its Magic9 phones, tablet and watch on Sept. 28, the same day HIMA launched the Zhijie RX SUV. Anthropic, AMD and Manus also made announcements.

Policy & Regulation · Large Language Models · Companies

Florida AG seeks injunction against OpenAI as company scraps GPT-6.1 Astra

Florida Attorney General James Uthmeier is seeking a court injunction to block OpenAI from advancing frontier models without independent oversight and to keep minors off its technology, as OpenAI scraps GPT-6.1 Astra over safety concerns.

Policy & Regulation · Large Language Models · Companies

Florida Attorney General Seeks Injunction to Curb OpenAI Frontier Model Development

Florida Attorney General James Uthmeier asked a state court to block OpenAI from advancing frontier models without third-party oversight and to keep minors from using its technology. OpenAI said it is cooperating with officials and has paused training its most capable models.

AI Chips & Compute · Large Language Models · Companies

AMD to Acquire World Labs for $8.2 Billion; Apple Fixes iPhone 18 Pro Face ID Bug

AMD agreed to buy Fei-Fei Li's World Labs in an $8.2 billion stock deal. Apple released iOS 27.0.1 to fix iPhone 18 Pro Face ID reboots, while Anthropic, SpaceX, Xiaomi and Manus also announced new models, launches and services on Sept. 28.

Companies · Large Language Models

OpenAI Shelves GPT-6.1 Astra Release After It Fails Internal Safety Tests

OpenAI shelved GPT-6.1 Astra after tests found it failed alignment checks and showed more deception than its predecessor.

Large Language Models · Products & Applications

Anthropic Adds Claude Sonnet 5.5 as a Faster, Cheaper Option Below Opus 5.5

Anthropic has introduced Claude Sonnet 5.5, a lower-cost, faster companion to Opus 5.5 that the company says improves substantially on Sonnet 5 while keeping the same low operating cost.

Policy & Regulation · Companies · Large Language Models

Tech executives including Zuckerberg and Amodei expected at Trump luncheon Tuesday

Anthropic CEO Dario Amodei and Meta CEO Mark Zuckerberg are among tech executives expected to join President Donald Trump and House Speaker Mike Johnson at a Tuesday luncheon tied to a broader White House event on AI, energy and space.

Companies · Products & Applications · Large Language Models

Meta Launches Enterprise AI Unit, Names CJ Desai to Lead It

Meta is launching Meta Enterprise Platform, a new enterprise AI business led by former MongoDB CEO CJ Desai. It will initially push Muse, Muse Code, Meta Business Agent and Muse APIs to companies and developers.

Companies · Large Language Models

Vercel Data Shows AI Model Churn Is Accelerating as Open-Weight Models Gain Ground

Vercel’s AI Gateway data show most AI tokens are spent on models less than four months old, open-weight models account for 56% of token volume, and average token prices fell 23.2% for a third straight month.

Companies · Large Language Models · Products & Applications

Anthropic's Claimed AI Lab Discovery Draws Challenge From Biologists

Anthropic says a Claude-powered biology lab found a previously uncatalogued genetic pattern, but biologists dispute that it is a true scientific discovery, and one Copenhagen researcher says his team found the pattern first.