Anthropic just dropped Opus 4.8... (WOAH)
Matthew Berman showed 4 AI tools in this video on . 2 are verified available in the EU. Each row below jumps to the exact second. Part of week 2026-22.
Tools shown in this video, with timestamps
Anthropic's new model, building on Opus 4.7 at the same price ($5/$25 per million tokens). It scores 69.2% on SWE-Bench Pro, a 5-point gain, leads on Humanity's Last Exam, and trails GPT-5.5 on Terminal Bench 2.1. Fast mode runs about 2.5x faster and is now cheaper.
Live demoService that lets agents publish anything to the web instantly via a public link. It was used to share the Opus 4.8-made game with stream viewers. It recently added a private storage feature.
Live demoResearch preview feature where Claude writes orchestration scripts that run tens to hundreds of parallel sub-agents. It verifies results adversarially and targets bug hunts, large migrations and audits. It can use substantially more tokens, and a new 'ultra code' setting sets effort to extra high. No benchmarks were given.
Anthropic says it plans to release a new class of model with higher intelligence than Opus in the coming weeks. It is currently in preview for cybersecurity work under Project Glasswing, pending stronger cyber safeguards.
