← OpenAI
Matt Wolfe
AI News: The Biggest Leap We've Seen This Year!
4 min read26 Apr 2026Worth watching
TL;DR
Matt Wolfe covers a packed week in AI, headlined by GPT-5.5 (now the top-ranked model on the Artificial Analysis intelligence index), ChatGPT Images 2.0 (beating Nano Banana on LM Arena), and Claude Design with animation capabilities. Multiple other models from Alibaba, Kimi, and Google also dropped, plus robotics news from a China half-marathon where four robots finished in under an hour.
Key points
1
GPT-5.5 is now the top-ranked model on the Artificial Analysis composite intelligence index, beating Claude Opus 4.7 and Gemini 3.1 Pro, and outperforms Anthropic's unreleased Mythos model on Terminal Bench (82.7% vs 82%)
2
ChatGPT Images 2.0 jumped to an LM Arena score of 1500, far ahead of Nano Banana at 1271, and can generate functional barcodes, dense text, and research-informed infographics
3
Claude Design lets users create animated presentations, prototypes, and websites with minimal prompting, though it defaults to a recognizable dark-mode aesthetic across all outputs
4
GPT-5.5 is priced at $5 per 1M input tokens and $30 per 1M output tokens (double GPT-5.4), but uses significantly fewer tokens to complete the same tasks, partially offsetting the cost increase
5
Anthropic's unreleased Mythos model was accessed by unauthorized users after Anthropic publicly marketed it as too powerful and scary to release, which Sam Altman compared to selling bomb shelters
Actionable insights
→
GPT-5.5 users on Plus, Pro, Business, or Enterprise can access the model now in ChatGPT and Codex, but the API is not yet available and it is not in Cursor or third-party coding tools yet
→
ChatGPT Images 2.0 is accessible directly in ChatGPT via the plus button; it works best for dense-text images, infographics, multi-language content, and magazine-style layouts
→
Claude Design is available at claude.ai for Pro, Max, Team, and Enterprise users and can generate basic After Effects-style animations from simple text prompts, useful for presentations and explainer videos
→
Qwen 3.6 27B is a fully open-source coding model from Alibaba, and Kimi K2.6 is an open-weight model that can run 300 parallel sub-agents and beats Opus 4.6 on several benchmarks
→
OpenAI released a Privacy Filter model on Hugging Face and GitHub that can mask PII locally without data leaving your machine, and ChatGPT for Clinicians is free for verified US clinicians
Notable quotes
“It is clearly incredible marketing to say, we have built a bomb. We are about to drop it on your head. We will sell you a bomb shelter for $100 million.”
“GPT-5.5 is a model that performs better on Terminal Bench than the model that Claude refuses to release because it is too scary.”
“It is getting to a point where I feel like image generation is fairly solved. Like, there is not much we cannot do with it.”
Worth watching?
✅
Worth watching the full video?
Worth watching if you want to see the live demos of GPT-5.5 personalization, the Images 2.0 barcode trick, and Claude Design animations in action — but the key specs and comparisons are all captured here.
Saved you some time? The creator still deserves a like.
Want this for your own channels?
Add the channels you follow. Every new video summarised and in your inbox the moment it drops. From £4/month.
Try it free