Gemini Carbon, NVIDIA CEO AFFAIR, Fable 5.2, Meta Muse Ember, GPT Image 2.6, Minimax H3: AI NEWS
Matthew Berman showed 11 AI tools in this video on . 5 are verified available in the EU. Each row below jumps to the exact second. Part of week 2026-46.
Tools shown in this video, with timestamps
Creator calls it the best model available, citing Artificial Analysis results where Fable 5.2 holds the top three spots on the Intelligence Index. It is also the most expensive, and burns through quota quickly.
Google's new frontier model, now available. The creator demos an Excel clone, a Fall Guys clone and a Rube Goldberg machine. He finds it good, cheaper than Fable 5.2, but well behind Fable in quality.
Live demoOpen 1.6T parameter multimodal model with 48B active parameters, with stronger visual understanding and strength in creative design and 3D tasks. Too large for most people to run locally.
US-made open-weights 65B dense LLM scoring 80.5% on Terminal Bench 2.1 versus other open models. It has a 1M token context, runs locally on a MacBook or DGX Spark, and costs $0.20/$0.60 per million tokens.
Open-weights small model scoring 31 on the Artificial Analysis Intelligence Index. It competes well with the larger DeepSeek V4.1 Flash on several benchmarks, and the demos include a SimCity-style game and a PC emulator.
Live demoMeta's smallest model. It sits on the Pareto frontier of cost against intelligence, at about a third of Claude Haiku's price. It beats Haiku and GPT-6 Luna on several benchmarks and ties Haiku 5.5 on the Intelligence Index. Suggested as a delegate model for subtasks.
New OpenAI image model that adds GIF generation and is about 50% faster. It is available in ChatGPT, Codex and the API in two versions, Sunburst for quality and Flare for speed.
Google's video generation model, now 30% cheaper and faster, with longer clips, in-conversation editing, more natural speech and more controllable shots.
ByteDance image model that the creator calls possibly the best image generator right now. He highlights hyper-realistic detail and consistent multi-turn editing.
Live demoCognition's fast, efficient coding model, post-trained from Inkling Small. It is meant to be used by sub-agents on delegated coding tasks.
Fully open-weights image generation model. It is 68 GB, with an 18 GB quantized version that most computers can run.