Claude 4.1 likely in internal testing as Anthropic prepares safety checks

Anthropic is testing Claude Opus 4.1, focusing on upgraded reasoning and planning. Safety checks via Neptune v4 red teaming suggest a near-term launch, targeting professionals using Claude for research, development, and automation tasks.

🗞 #claude

DeepMind debuts Genie 3 to turn text prompts into playable 3D worlds

DeepMind's Genie 3 converts text prompts into playable 3D environments at 720p/24fps, offering multi-minute playtime and new in-run changes. It supports AI training and lowers simulation costs, advancing DeepMind’s world-as-simulator strategy.

🗞 #ai

Anthropic rolls out Claude Opus 4.1 with 74.5% SWE-bench score

Anthropic has released Claude Opus 4.1, showing improved coding accuracy and task handling while maintaining pricing and compatibility. It is available on Anthropic API, Amazon Bedrock, and Google Cloud Vertex AI platforms.

🗞 #claude

OpenAI debuts GPT-OSS with 120B and 20B parameter models

OpenAI releases GPT-OSS, its open-weight language models with 120B and 20B parameters, rivaling closed models in performance. Distributed under Apache 2.0, they are designed for local deployment and meet stringent safety and benchmark standards.

🗞 #chatgpt

Gemini can now turn sketches and photos into custom Storybooks

Google's Gemini app now includes Storybook, a tool that turns user prompts into illustrated, narrated 10-page stories in over 45 languages. It supports photo integration and exports, targeting parents and educators for personalized storytelling.

🗞 #gemini

Google launches Gemini CLI GitHub Actions in public beta for developers

Google’s Gemini CLI GitHub Action installs an AI agent into repositories, enabling asynchronous bug labeling, fix drafting, and release note generation using up to one million tokens of context, with support for secure, scheduled task execution.

🗞 #gemini

OpenAI's GPT-5: What we know so far about August 7 release

OpenAI will unveil GPT-5, along with GPT-5 Mini and Nano variants, aimed at diverse developer and enterprise needs. Partners like Microsoft are already integrating the models, which show marked gains in reasoning and context processing.

🗞 #chatgpt

Students get free Google AI Pro with new Guided Learning mode in Gemini

Google is offering students in five countries free access to its AI-powered Gemini study tools and a year of Google AI Pro, featuring the Gemini 2.5 Pro model and other premium resources as part of a global education initiative.

🗞 #gemini

Leaked details reveal GPT-5 Pro with research-level AI for ChatGPT

OpenAI is set to release GPT-5 with a redesigned ChatGPT interface and multiple model versions, including a research-grade Pro variant. Access will be tiered to align capabilities with user subnoscriptions and enterprise needs.

🗞 #chatgpt

Perplexity preps platform for possible GPT-5 rollout

Clues from Perplexity hint that GPT-5’s release may be imminent, with early signs of rollout through select partners. Evidence suggests multiple internal variants and auto-routing capabilities, though official confirmation is still pending.

🗞 #perplexity

Microsoft tests Copilot Creator Gallery for Sora-style AI image and video feed

Microsoft is testing a new internal Copilot feature called “Creator Gallery,” a feed of AI-generated images and videos designed for creators to view, like, remix, and access prompts, aligning with its broader media content strategy.

🗞 #microsoftcopilot

India may get new Google AI plan Helium at ₹250 launch price

Google is testing a ₹500 India-specific Gemini subnoscription plan, Helium, launching at ₹250 with access to premium tools like Veo 3 and Deep Research. The move targets early adoption amid growing AI competition in the price-sensitive market.

🗞 #gemini

Perplexity works on remote MCP integrations for Notion and Linear

Perplexity is developing remote MCP support with integrations for tools like Linear and Notion, aiming to embed them into its search workflow. This shift aligns it with AI assistants operating across multiple apps and data sources.

🗞 #perplexity

Claude Sonnet 4 offers 1M token context window for large-scale enterprise tasks

Anthropic has expanded Claude Sonnet 4's context window to 1 million tokens via API and Amazon Bedrock, targeting enterprise-grade tasks like large codebases and contract analysis. Access starts with premium tiers, with broader rollout planned.

🗞 #claude

Google develops Magic View for NotebookLM as possible new way to visualize notebook data

Google is testing a new NotebookLM feature called Magic View, suggesting a visual or interactive way to explore notebook content. A broader rollout of video overviews with language options also reflects its push for accessible AI-driven tools.

🗞 #notebooklm

Google rolls out Temporary Chat in Gemini and new personalisation controls

Google introduces Temporary Chat for Gemini, offering a non-personalized session with stricter data limits and expanded privacy controls. The rollout starts on the 2.5 Pro model, focusing on transparency and user control in personalization settings.

🗞 #gemini

Google readies image generation update ahead of August 20 Pixel event

Google is preparing to launch new AI-driven features across its ecosystem, including GEMPIX image generation, the lightweight Nano Banana model, and Magic View—all likely tied to the upcoming Pixel 10 announcement on August 20.

🗞 #gemini

Google rolls out Gemma 3 270M multimodal model for phones and edge inference

Google's Gemma 3 270M model offers efficient, task-specific performance with INT4 support, low battery usage, and multilingual capabilities. It's optimized for fine-tuning and structured tasks, with broad deployment and tooling support.

🗞 #gemini

OpenAI readies ChatGPT to power its upcoming AI browser

OpenAI is developing a Chromium-based browser, codenamed "Aura," with a Mac-first rollout. It includes a custom new tab page and integrated cloud agent mode, allowing direct web access without repeated logins, similar to Perplexity's approach.

🗞 #chatgpt

Google develops Projects feature for Gemini

Google is developing a "projects" feature for Gemini, allowing users to manage files, set project instructions, and access research tools within AI conversations. No release date is confirmed, and the feature remains under development.

🗞 #gemini

Gempix surfaces in Whisk as Google tests precise reference modes

Gempix has surfaced in both Gemini and Whisk, indicating upcoming support for reference-based image editing using Imagen 4. Google may introduce these AI tools at its August 20 event, initially for Pixel devices and Labs experiments.

🗞 #aitestkitchen