Daily AI News

Anthropic Releases Claude Fable 5.1 and Claude Mythos 5.1: 52.6% on Terminal-Bench-Science and 75% Cheaper Cache Reads

Anthropic launched two updated versions of its Claude series. Fable 5.1 is generally available via API, AWS, Google Cloud and Microsoft Foundry, while Mythos 5.1 remains limited to vetted organisations under Project Glasswing. The upgrades improve benchmark performance (52.6% on Terminal‑Bench‑Science) and cut cache‑read costs by 75%.

Read source →

Meta Superintelligence Labs Releases Muse Voice Transcribe: One Real-Time Model for Streaming ASR, Diarization, and Endpointing

Meta unveiled Muse Voice Transcribe, a single model that simultaneously performs automatic speech recognition, speaker diarization and endpoint detection in real time, removing the latency of traditional multi‑system pipelines. The unified approach simplifies production voice stacks and reduces failure points.

Read source →

Path to Astra

OpenAI announced Astra, its first model that meets the Critical cybersecurity capability threshold of the Preparedness Framework, incorporating stronger release safeguards. Astra is positioned as a benchmark for secure, frontier AI deployments.

Read source →

Introducing agentic video understanding with Gemini

Google added agentic video‑understanding capabilities to its Gemini family, delivering higher accuracy and lower computational cost for video tasks while reducing token usage. The feature enables developers to build more efficient video AI applications at scale.

Read source →

Perplexity Releases Hybrid Compute on Mac: Cloud Agents Orchestrate Down to a Local Model, Gated On Device

Perplexity launched a hybrid‑compute solution for macOS that lets cloud‑based agents coordinate with a locally gated model, keeping sensitive data on‑device while still leveraging powerful AI. The system dynamically splits workloads between cloud and device for privacy‑preserving performance.

Read source →

Try Google Pics: Easy image creation and editing in Google Workspace

Google introduced Google Pics, an image‑generation and editing tool built on its Nano Banana model, now embedded in Google Workspace apps like Docs, Slides and Chat. Users can create and modify visuals directly within their productivity workflow.

Read source →

Amazon Alexa can now alert you when something new might tempt you to shop

Amazon added an “Update Me When” feature to Alexa that sends personalized alerts about product launches, tours, books, shows and other events that could trigger a purchase. The alerts aim to give users more control over impulse‑shopping triggers.

Read source →

Nvidia’s controversial DLSS 5 arrives September 3rd and requires serious GPU horsepower

Nvidia announced DLSS 5, its next‑generation AI upscaling technology, slated for a September 3 release and limited to the RTX 50‑series desktop GPUs due to its high compute demands. The update promises more realistic rendering but has sparked debate over performance trade‑offs.

Read source →

AfterQuery reportedly becomes Y Combinator’s fastest-ever unicorn, now valued at $3.2B

AI model‑training startup AfterQuery raised a fresh round that pushes its valuation to $3.2 billion, making it Y Combinator’s quickest unicorn after only five months since its Series A. The company focuses on accelerating large‑scale model training for enterprises.

Read source →

Apple accuses OpenAI of destroying evidence

Apple filed a lawsuit claiming OpenAI deliberately destroyed evidence in a trade‑secret dispute and is seeking expedited discovery. The case adds to a growing wave of legal battles over AI‑related intellectual property and data practices.

Read source →