Actualités de l'IA

Pace the Frontier: What AI Labs Actually Committed To

On 12 September 2026 Anthropic CEO Dario Amodei published an essay arguing that the AI industry should deliberately slow down, and within hours the heads of OpenAI…

iPhone 18 Pro AI Features: Every One, and Which Need the New Phone

Apple gave AI top billing at the iPhone 18 Pro launch on 9 September 2026, and the A20 Pro doubles the Neural Engine to 32 cores. Yet almost every AI feature Apple…

iPhone Duo: Every AI Feature in Apple's First Foldable

Apple announced the iPhone Duo on 9 September 2026, its first foldable iPhone, at $1,999 for 256GB. It runs the same A20 Pro chip as the iPhone 18 Pro, so every AI…

Apple Audio Intelligence: What Your Apple Watch Actually Hears

Apple announced Audio Intelligence on 9 September 2026, a set of four Apple Watch features that includes a tool which summarises the conversations you have during…

ChatGPT Images 2.5 est là : génération plus rapide, retouches par commentaire et deux nouveaux modèles d'API

OpenAI a lancé ChatGPT Images 2.5 le 8/9/2026, avec jusqu'à 50 % de latence en moins, l'édition par commentaires, l'outil Sketch et deux nouveaux modèles d'API. Ce…

When Will AGI Happen? Every Prediction, Tracked and Scored

Every frontier lab CEO has named a year for AGI, and several of those years have already gone by. This is the full board: who predicted what, the date they said it…

iPhone Buying Guide 2026: Buy Now, or Wait for the iPhone 18?

Apple’s event has been and gone. The iPhone 18 Pro starts at $1,199, the foldable is called the iPhone Duo, and the phone actually named the iPhone 18 is not coming…

GPT-6 Astra est là : benchmarks, prix et ce que les chiffres ne disent pas

OpenAI a lancé GPT-6 Astra le 3 septembre 2026 et le présente comme l’état de l’art en utilisation de l’ordinateur, ingénierie logicielle et science. C’est le…

Muse Spark 1.3: What Meta’s Own Benchmark Chart Actually Shows

Meta released Muse Spark 1.3 on September 2, 2026, its fourth Muse Spark in five months. It wins every coding and long-context row on Meta’s own chart and loses…

Gemini 3.8 Flash: Benchmarks, Pricing and What the Headlines Got Wrong

Google made Gemini 3.8 Flash generally available on September 2, 2026, three weeks after 3.7 Flash. It wins 8 of the 14 rows in Google's own comparison table and…

Claude Fable 5.1 and Mythos 5.1: One Model, Two Sets of Safeguards

Anthropic released Claude Fable 5.1 and Claude Mythos 5.1 on September 1, 2026. They are the same model with two different safety layers, and the system card calls…

Tencent Hy4 Preview: The Biggest Apache 2.0 Model Yet, and What It Actually Scores

Tencent released Hy4 preview on August 28, 2026 and open-sourced it the same day, 770 billion parameters under Apache 2.0. That makes it the largest permissively…

OpenAI AGI in 2026: What Altman Actually Promised

Sam Altman told TIME that OpenAI will have an internal system he would call AGI by the end of 2026. The claim is real, the quote is accurate, and it means…

GLM 5.3 Flash: Z.ai Ships Ox Alpha as a 320B Open Model

Z.ai released GLM 5.3 Flash on August 26, 2026, and confirmed it is the stealth model that had been running on OpenRouter as Ox Alpha. It is a 320B model with 18B…

Are AI Companies Destroying Books? What the Court Files Actually Show

Court documents unsealed in January 2026 show Anthropic bought millions of used print books, cut off the spines, scanned the pages and threw the originals away. In…

GLM 5.3: What Z.ai Shipped, and the Weights It Held Back

Update, August 26, 2026: the two-week deadline has passed and GLM 5.3’s weights are still not out. Z.ai released a different model instead, the 320B multimodal GLM…

Gemini 3.7 Flash: Pricing, Benchmarks and the Catch Google Buried

Google released Gemini 3.7 Flash on August 13, 2026, just 23 days after Gemini 3.6 Flash. Coding scores jumped hard and the price halved to $0.75 per million input…

AI Models Discontinued in 2026: The Running List

OpenAI, Anthropic and Google retired dozens of AI models in 2026, and more shutdowns land every month. Here is the running list of what is already gone, what…

Grok 4.6: What's New, Benchmarks and Pricing

SpaceXAI released Grok 4.6 on 12 August 2026. It scores 61 on the Artificial Analysis Intelligence Index, ties GPT-5.6 Sol, and costs 60 percent less than the…

GPT-5.6-Cyber: OpenAI’s Hacking Model, and Why You Cannot Use It

OpenAI released GPT-5.6-Cyber on August 10, 2026, a model trained to find zero-days and build exploit chains, tuned to refuse far less than the ChatGPT you use…

What Is Grok Bot? SpaceXAI’s AI Agent App for Mac and iPhone

SpaceXAI launched Grok Bot on 11 August 2026, and it is not a chatbot. Each account gets a persistent cloud computer that AI agents use to sign into your apps with…

AI Safety Incidents in 2026: The Running List

The most serious AI safety incidents of 2026 did not happen in the wild. They happened inside the labs’ own testing, and the UK government has now published the…

Muse Spark 1.2 and Muse Code: Meta’s Coding Agent for Mac

Meta shipped Muse Spark 1.2 on 5 August 2026 alongside Muse Code, its first terminal coding agent, and the launch charts tell a more honest story than the headlines…

Apple Sues OpenAI: What It Means for Siri and Your iPhone

On Monday 3 August 2026, Apple asked a federal judge to bar OpenAI and two former Apple employees from accessing, using or disclosing its confidential information…

Des modèles IA ont piraté Hugging Face pour tricher à leur propre test

Le 21 juillet 2026, OpenAI a confirmé quelque chose qu'aucun laboratoire IA n'avait jamais dû admettre auparavant. Deux de ses propres modèles ont échappé à un…

Claude Opus 5 Review: Pricing, Benchmarks, and the New Effort Setting

Claude Opus 5 is here. Anthropic released the model on July 24, 2026, and the headline is hard to ignore: it delivers close to Claude Fable 5 intelligence at Opus…

Sakana AI’s Fugu-Ultra v1.1 Is Here: Better Benchmarks, Same Price

Sakana AI has released Fugu-Ultra v1.1, an upgrade to its frontier orchestration model that the lab says gains up to 7.9 points over v1.0 across every benchmark it…

Gemini 3.6 Flash Is Here: Pricing, Benchmarks and What Changed

Google released Gemini 3.6 Flash on July 21, 2026, and the headline number is not a benchmark. It is the price. Output tokens dropped to $7.50 per million, down…

Is Codex Wearing Out Your Mac’s SSD? The Codex SSD Bug Explained

One developer measured roughly 37 TB written to their SSD in just 21 days of running OpenAI’s Codex, which extrapolates to about 640 TB per year. That is a problem…

Qwen 3.8: Alibaba’s 2.4T Model Tested Against the Fable 5 Claim

Qwen 3.8 is no longer a preview with a marketing claim attached to it. Alibaba made Qwen3.8-Max generally available on August 3, 2026, and this time the model…