EDITORIAL ARCHIVE

AI news

News, tests and analysis to understand what changed, what is hype and what your business can already use.

290stories on the radar

Latest from the radar

Page 7 / 25
Releases & Models#1186
73 Releases & Models

Multiverse launches Quasar 438B and sells a European leader title

Spain's Multiverse Computing published Quasar 438B, a 438 billion parameter model for agents and coding. Artificial Analysis scores it 43, above Mistral Medium 3.5. The cut is English and Spanish, a closed API, no Portuguese and no open weights.

3 min read
Releases & Models#1175
74 Releases & Models

Gemini turns on agentic video understanding

Google turned on agentic video understanding in Gemini 3.7, 3.6 Flash and 3.5 Flash-Lite. Vendor numbers: up to 88% fewer tokens and 66% lower cost. Not Gemini 3.8.

1 min read
Releases & Models#1173
75 Releases & Models

CONFIRMED: xAI launches Grok 4.7 for long-running tasks at $2 per million input tokens

xAI announced Grok 4.7 on September 21, with a larger base model, longer reinforcement training, and a focus on coding and knowledge work. Independent evaluations published soon after place the model mid-pack on the Artificial Analysis Intelligence Index, behind Claude Fable 5.1 and GPT-6 on agentic coding benchmarks such as Terminal-Bench.

2 min read
Releases & Models#1171
76 Releases & Models

Qwen3.8-Max-0902 gets independent confirmation and reaches OpenRouter

CONFIRMED: Qwen3.8-Max-0902 keeps the same pricing and context window, received independent validation on Code Arena WebDev, and is scheduled to become the Qwen3.8 Max route on OpenRouter on September 5.

1 min read
Releases & Models#1169
77 Releases & Models

DeepSeek open-sources V4-Flash-Vision-Exp, the V4 family's first eyes

DeepSeek published DeepSeek-V4-Flash-Vision-Exp on Hugging Face, the first experimental multimodal model in the V4 family, at 305B parameters under an MIT license. The pitch is not picture captioning for humans: it is vision for agents that read screens, charts and UIs and then call tools.

1 min read
Releases & Models#1167
78 Releases & Models

ChatGPT now connects Epic records and official healthcare sources

On September 1, 2026 OpenAI launched an Epic EHR integration and a Healthcare Public Data plugin inside ChatGPT for Healthcare, with read-only access to authorized patient context and nine official public sources.

1 min read
Releases & Models#1165
79 Releases & Models

Google launches Gemini 3.8 Flash and keeps Cyber for defenders

Google confirmed Gemini 3.8 Flash at 3.7 introductory pricing and gated 3.8 Flash Cyber behind Fairwind. This morning's ID is now a product.

3 min read
Releases & Models#1163
80 Releases & Models

Meta launches Muse Voice Transcribe, Superintelligence Labs' first real-time ear

Meta Superintelligence Labs launched Muse Voice Transcribe on September 1, 2026, its first real-time audio perception model: streaming ASR, diarization for 20-plus speakers and endpointing in one Muse Spark model, live on the Meta Model API, Meta AI for Mac and Muse Code.

2 min read
Tools#1161
81 Tools

AMD Launches ROCm 10 With AI-Native ROCm.AI Toolkit

AMD announced ROCm 10 on August 27, 2026, the latest version of its open software platform for GPU and artificial intelligence computing, marking ten years since the stack's debut as the main rival to Nvidia's CUDA. According to reports from StorageReview, VideoCardz, HPCwire, WCCFTech and AMD's own website, the release introduces ROCm.AI, a toolkit described as AI native that combines the agentic ROCm Hyperloom system, AMD Skills and the ROCm CLI. AMD states that a system configured with ROCm.AI delivers, on average, 3.3 times more inference performance and 2.4 times more training performance compared with ROCm 7, in tests run on a platform with eight Instinct MI355X GPUs using models such as GLM-5, Kimi K2.5 and DeepSeek-R1-0528. The move is widely read as another chapter in AMD's fight against Nvidia for control of AI infrastructure software.

3 min read
Tools#1160
82 Tools

Perplexity Launches Hybrid Mode for Mac, Blending Local AI With the Cloud

Perplexity AI began rolling out Hybrid mode for its Mac app on August 31, 2026, a feature that lets AI models run locally on the user's own computer, such as Google's Gemma 4 and Alibaba's Qwen3.6, to handle subtasks while Perplexity's cloud continues orchestrating the rest of the work. The company also offers its own proprietary model, aimed at machines with 32 GB of RAM, while Gemma 4 is recommended for Macs with 16 GB. The core idea is to combine the speed, privacy and zero per token API cost of local processing with cloud level orchestration for more complex tasks. The launch lands in the same month Perplexity expanded access to models including GPT-5.6, Grok 4.6, DeepSeek V4 Pro, Kimi K3 and Nvidia Nemotron 3.5 Lightning, and rolled out Computer in Email, reinforcing an increasingly hybrid, multi model platform strategy.

3 min read
Market & Business#1159
83 Market & Business

US Adds ChatGPT and Grok to GenAI.mil, But Grok's Safety Comes Into Question

On Monday, August 31, 2026, the US Department of War, the new name for the former Department of Defense, launched two new artificial intelligence tools on the GenAI.mil platform: ChatGPT Mil, a customized version of OpenAI's ChatGPT, and Grok for Government, from xAI, joining Google Gemini, which was already available on the portal. According to reporting from TechCrunch, DefenseScoop, The Hill, Hoodline, Android Headlines and Aroged, GenAI.mil is a centralized, secure portal launched in 2025 to give Department personnel access to frontier commercial AI models without routing sensitive government data through ordinary consumer channels. More than 3 million service members and civilians have potential access to the platform, and more than 1.7 million unique users have already been onboarded. ChatGPT Mil, accredited to handle Controlled Unclassified Information at Impact Level 5, offers chat, files, projects and custom GPTs to draft procurement materials, summarize policy documents and generate internal reports. Grok's arrival, however, has drawn controversy: according to TechTimes, internal engineers flagged that there is no reliable fix for the model's risk of generating child sexual abuse material, reviving questions about the safety criteria behind the US government's AI adoption.

4 min read
Market & Business#1158
84 Market & Business

ChatGPT Ads Tops $1 Billion in Annualized Revenue in Under 200 Days

OpenAI announced on Monday, August 31, 2026, that ChatGPT Ads, its advertising operation inside the chatbot, surpassed $1 billion in annualized revenue run rate in less than 200 days since launch. According to reporting from Yahoo Finance, PYMNTS and Adweek, the platform is now available in more than 40 countries and serves tens of thousands of active advertisers. Digiday and TechTimes note that, alongside the milestone, OpenAI opened self-serve access through Ads Manager to advertisers in India, Europe, the Middle East and North Africa. The company is targeting $2.5 billion in advertising revenue this year and internally projects an annualized revenue rate above $40 billion, nearly double the pace it was running at by the end of 2025. The move comes just ahead of a potential OpenAI initial public offering (IPO), with advertising treated as a central piece of the company's revenue diversification strategy.

3 min read
WhatsApp