In partnership with

Welcome back!

OpenAI just dropped ChatGPT Images, and it’s setting a high benchmark as it reportedly outperforms Google Nano Banana. The race between OpenAI and Google is heating up. Let’s unpack…

Today’s Summary:

• 🚀 ChatGPT Image 1.5 challenges Google Nano Banana

• 🎧 Meta releases SAM Audio editor

• 💻 Mistral launches Devstral 2 for local vibe coding

• 🗣️ Google updates Gemini native audio

• ⚡ NVIDIA unveils Nemotron 3 open agent models

• 🔍 Gemini Deep Research Agent gets major upgrade

• 🛠️ 2 new tools

TOP STORY

Did ChatGPT Image dethrone Google’s Nano Banana?

The Summary: OpenAI has rolled out GPT Image 1.5, a faster and more accurate image generation model now live in ChatGPT. It delivers edits up to 4x quicker, while keeping lighting and facial details consistent. LMArena benchmarks show GPT Image 1.5 outperforming Google Nano Banana Pro, though early user feedback suggests Nano Banana still leads in real use.

Key details:

• ChatGPT Image 1.5 preserves facial consistency, logos, and brand color tones across edits

• Handles complex prompts with strong instruction following

• 20% lower API pricing than previous version

• Rolling out globally across ChatGPT and API

Why it matters: This launch is part of OpenAI’s recent “code red” push to catch and outpace Google. ChatGPT Images merges GPT-5.2 reasoning with image generation, creating a unified reasoning image engine that achieves the impressive feat of catching up with Google Nano Banana.

ChatGPT Images FAQ

FROM OUR PARTNERS

Turn Any Task into a How-To Video

Effortless Tutorial Video Creation with Guidde

Transform your team’s static training materials into dynamic, engaging video guides with Guidde.

Here’s what you’ll love about Guidde:

1️⃣ Easy to Create: Turn PDFs or manuals into stunning video tutorials with a single click.
2️⃣ Easy to Update: Update video content in seconds to keep your training materials relevant.
3️⃣ Easy to Localize: Generate multilingual guides to ensure accessibility for global teams.

Empower your teammates with interactive learning.

And the best part? The browser extension is 100% free.

Check out Guidde

META

Meta releases SAM Audio for sound editing

The Summary: Meta has launched SAM Audio, an AI model that isolates specific sounds from complex recordings using text, visual, or time-based prompts. The model separates vocals, instruments, or background noise and runs faster than real time.

Key details:

• Built for creators, podcasters, musicians, film editors, and anyone who wants to manipulate audio without complex tools

• Users can mute a barking dog, isolate vocals from a song, or pull a sound effect from noisy footage by describing it in text or clicking on it in a video

• Available free through the Segment Anything Playground for anyone to test and remix audio interactively

Why it matters: This model brings to audio the same editing simplicity that already exists for text and images. Traditional tools let users slice or filter audio waveforms, but SAM Audio treats it as structured content that can be referenced by meaning, enabling text commands such as “remove traffic noise” or “extract vocals”.

FROM OUR PARTNERS

A Smarter Way to Read the News

Seeking impartial news? Meet 1440.

Every day, 3.5 million readers turn to 1440 for their factual news. We sift through 100+ sources to bring you a complete summary of politics, global events, business, and culture, all in a brief 5-minute email. Enjoy an impartial news experience.

Join for free today!

MISTRAL AI

Mistral Launches Devstral 2 for local vibe coding

The Summary: Mistral AI has released Devstral 2, an open-source family of coding models featuring 123B and 24B parameter versions. The flagship model achieves 72.2% on SWE-Bench Verified (#2 open source model after DeepSeek-V3.2) and runs on four H100 GPUs, while the still impressive smaller variant runs locally.

Key details:

• Devstral Small 2 (24B) runs on a single RTX 4090 or CPU-only setups, with a 256K token context

• 5x smaller than DeepSeek V3.2 and 8x smaller than Kimi K2

• Mistral Vibe CLI provides project-aware scanning, version control integration, and programmable automation via ACP protocol

Why it matters: Devstral 2 is a strong open coding model with agentic autonomy. The 256K context and tool-calling feedback loop let it write, test, and repair code autonomously. Paired with the Vibe CLI, it enables running an entire development workflow locally, without depending on Anthropic / OpenAI / Google API services.

QUICK NEWS

Quick news

• Gemini Deep Research Agent gets a major upgrade

• Google updates Gemini Flash Native Audio for live voice agents

• NVIDIA Nemotron 3 fast, open agentic models

TOOLS

🥇 New tools

• PlanEat - Turns your health goals into menus and grocery lists

• CC - AI agent for your daily briefing