Welcome back!

According to a new report, OpenAI’s Strawberry is set to launch in two weeks, ahead of schedule. The new reasoning-focused model will introduce a longer “thinking pause” before delivering responses. The countdown begins…

Today’s Summary:

• 🍓 OpenAI’s Strawberry set to launch in 2 weeks

• 📱 iPhone 16 new Apple Intelligence features

• 🧬 Open-source Chai-1 takes on AlphaFold 3

• 🖥️ DeepSeek-V2.5 excels in coding tasks

• 🩺 Harrison.rad.1 AI for radiology exams

• 🖼️ Mistral releases Pixtral 12B multimodal AI

• 🛠️ 2 new tools

TOP STORY

OpenAI’s Strawberry to be released in two weeks

The Summary: OpenAI’s Strawberry, the new reasoning-focused ChatGPT model, is set to launch in two weeks, according to a new report. This timeline is earlier than the previously expected fall release. Strawberry is notable for its “thinking pause” before responding, lasting 10 to 20 seconds. The exact integration method with ChatGPT remains unclear, but it's expected to be a standalone offering within the service.

Key details:

• Strawberry to be released as part of ChatGPT in two weeks

• Introduces a 10-20 second thinking pause before responses

• Initial version will only support text input and output, not images

• Capable of "deep research" and solving multi-step problems autonomously

Why it matters: Strawberry’s reasoning-focused design could enable more advanced workflows, handling complex tasks and performing autonomous research. The accelerated release suggests OpenAI's push to maintain a competitive edge.

APPLE

New iPhone 16 brings Apple Intelligence to your pocket

The Summary: Apple has unveiled the iPhone 16 lineup, featuring the new Apple Intelligence AI system. Key upgrades include visual search capabilities, an improved Siri assistant, and AI features integrated across iOS apps. The new 3 nm A18 Pro chip enables on-device AI processing, while a partnership with OpenAI brings ChatGPT integration.

Key details:

• Camera Control button on iPhone 16 Pro enables visual search and AI-powered image recognition

• Siri gets a major upgrade with natural language understanding and ability to work across apps

• watchOS 11 introduces an AI-powered Translate app

• AirPods users can interact with Siri by nodding or shaking their heads

• Apple Intelligence will roll out gradually on iPhone 15 Pro and iPhone 16, starting with US English users in October

• AI-focused macOS Sequoia will be released on September 16

Watch the video keynote (forward to 0:47:00 for the AI section)

BIOTECH AI

Open-source Chai-1 rivals Google AlphaFold 3 in protein structure prediction

The Summary: Chai Discovery has released Chai-1, a multimodal AI model for molecular structure prediction. It matches state-of-the-art performance across various drug discovery tasks, including protein-ligand interactions and multimer predictions. The model is available for free, via a web interface and as open-source code for non-commercial applications.

Key details:

• Chai Discovery raised $30M from Thrive Capital and OpenAI to challenge Google DeepMind’s drug discovery models

• Chai-1 matches AlphaFold 3's accuracy in predicting protein structures

• Unlike AlphaFold 3, Chai-1 is available for unlimited use, including for drug companies

• Requires less genetic data than AlphaFold 3 and can use lab data for improved predictions

Why it matters: While AlphaFold 3 is kept under wraps by Google, Chai-1 democratizes access to AI tools for molecular modeling, which could accelerate drug discovery efforts across academia and industry.

Technical report

QUICK NEWS

Quick news

• DeepSeek-V2.5 best open-source model for coding tasks

• Harrison.rad.1 AI model for radiology examination

• Mistral releases Pixtral 12B open-source multimodal AI

TOOLS

🥇 New tools

• Scispace AI Academic Writer - Write like a scientist

• Trupeer - Turn screen recordings to product videos & docs