Hi tech & AI enthusiasts, this week's highlights:
OpenAI's GPT-5.5 Just Took the Lead Back From Claude
ChatGPT Images Can Finally Spell
DeepSeek's Free AI Reads a Whole Book at Once
The Lab: How to Create Infographics With ChatGPT Images 2.0
AI Image of the Week for: Lifestyle and home decor brand campaigns
Let’s get into it.
▲ AI MODELS
OpenAI launched GPT-5.5, pitched as 'a new class of intelligence for real work.' It now leads the public model benchmarks, ending months of Anthropic dominance.
What you need to know
Beats Claude Opus 4.7 across most benchmarks: GPT-5.5 hits 82.7% on Terminal-Bench 2.0 against Claude's 69.4%, and leads on agentic coding, knowledge work, and long-horizon reasoning evals.
Built for messy, multi-step tasks: You can hand it an ambiguous, ongoing task and trust it to plan, use tools, check its work, and keep going without constant nudging.
Same speed, fewer tokens: It matches GPT-5.4 speed while using significantly fewer tokens to complete the same Codex tasks.
Key Ideas
▲ Why it matters: After months of Anthropic leading on quality, the momentum has flipped back to OpenAI. With Claude users complaining about rate limits and slower responses, GPT-5.5 lands at the right time. For business users, this means the smartest model is once again sitting inside ChatGPT, the tool most teams already pay for.
▲ AI TOOLS
OpenAI released ChatGPT Images 2.0 (model name gpt-image-2), now available to all ChatGPT users and via the API. It can finally render words inside images correctly.
What you need to know
Words come out right: Around 99% character accuracy, finally usable for posters, infographics, menus, and ad creative.
Multilingual support works: Japanese, Korean, Chinese, Hindi, and Bengali all render cleanly. Useful for any business marketing across regions.
It plans before it draws: The model researches the topic, plans the layout, then generates. You can request up to 8 versions in one go at 2K resolution, with characters and products consistent across all of them.
Key Idea
▲ Why it matters: Until last week, AI images were unusable for anything with large amounts of text on the page, which ruled out most marketing assets. With this update, you can ship branded social tiles, infographics, and pitch deck visuals from a prompt. The tutorial below walks through how.
▲ AI MODELS
DeepSeek released V4, a free, open-source AI that holds up to 1 million words in a single chat. That is roughly seven Harry Potter books.
What you need to know
Two free versions: V4-Pro for top performance and V4-Flash for fast, cheaper responses. Both are open-source, so anyone can run them on their own infrastructure.
Drop in huge documents: A year of contracts, a stack of customer reviews, a full book. It reads the whole thing without losing context.
Performance rivals the paid leaders: V4-Pro trades blows with the top models from OpenAI and Anthropic, and ranks just behind Gemini 3.1 Pro on world knowledge.
Key Idea
▲ Why it matters: A 1M context window used to be a premium feature locked behind expensive plans. DeepSeek just made it the new floor and gave it away free. For any business that handles long documents like legal, finance, customer research, this model is now viable cost saving option.
▲ EVEN MORE NEWS
Other big ideas from the past week.
▲ ChatGPT Now Runs Workflows for Your Team: OpenAI launched workspace agents for Business, Enterprise, and Edu plans. Build an agent once (lead qualification, weekly metrics, vendor screening) and it runs on its own across Slack and ChatGPT.
▲ Claude Agents Now Remember What They Learn: Anthropic added memory to its Managed Agents. Rakuten reports a 97% drop in first-pass errors after switching, because the agent stops repeating the same mistakes.
▲ OpenAI Open-Sourced Symphony, Its Coding Team Manager: OpenAI released Symphony, a free spec that turns Linear into a control plane for AI coding agents. Some teams reported a 500% increase in pull requests landed.
The Lab
▲ AI EDUCATION
Now that ChatGPT can render text correctly inside images, you can use it to make branded infographics great for personal or business brands. I’ve walked through a framework, with examples for construction firms and digital marketers.
What you'll learn:
How to swap the subject to fit your niche
How to push for annotated callouts instead of generic sections
How to refine the result with simple follow-up prompts
Where to use the finished image (LinkedIn, blog headers, pitch decks, lead magnets)
🔬 Interested in learning about AI for business? Join The Lab waitlist.
▲ AI IMAGE SHOWCASE
Create this type of image with Midjourney
a neutral toned bedroom with warm lighting. a bed with a beige upholstered headboard and light beige bedding with a brown throw blanket is prominent. a textured pillow is on the bed. on the left nightstand, there's a brass lamp and a candle. a spherical glass pendant light hangs overhead. floor length beige curtains cover the wall behind the bed. on the right, a black arched mirror stands next to a framed line drawing of a face. a black minimalist lamp, potted plants, and candles are on the right nightstand. a striped beige rug is on the wooden floor.Image Usage Suggestion: Interior design portfolios, home staging marketing, real estate listings, lifestyle and home decor brand campaigns, hotel and short-term rental promotions, or any content that needs a calm, modern, neutral aesthetic.
💌 Reply to this email with your AI image generation or suggestion of what you’d like to see and I'll feature it in a future newsletter.
▲ SHARE YOUR THOUGHTS
Help shape the content you see here by giving feedback
Have specific feedback or want to get in touch? Reply to this email and I’ll get back to you.
Know someone who’d love this newsletter? Forward it to a friend and have them sign up here.
Thanks for reading, until next time.
Stay curious,
Matt Lok, Editor
Brought to you by Metalabs. Digital marketing consulting specialising in AI.