Google rolls out Genie 4 world model to trusted testers
A new benchmark called Realtime‑Bench tests AI on 50 interactive game and web tasks that require real‑time decisions. The test measures three qualities—temporal state awareness, adaptive deliberation, and time‑aware action selection—under three clock settings. GPT‑6 Astra tops the wall‑clock evaluation with a 36 % pass rate, while models reach 70‑80 % under a paused clock.
Key points
- Realtime‑Bench has 50 tasks, 36 % pass rate for GPT‑6 Astra on wall‑clock
- Genie 4 adds real‑time modification, physics simulation, sound modeling, and complex event simulation
- Metis proposes an explicit epistemic state to lower hallucinations in AI models
Google DeepMind announced Genie 4, a world‑model that adds real‑time modification, physics and human action simulation, native sound generation, and complex event simulation. The model is currently a research preview for trusted testers and will later be available to Google AI Ultra subscribers.
Metis, a startup, released a paper proposing an explicit epistemic state to reduce hallucinations, arguing that adding a reasoning core could make AI more reliable.
Model pages: GPT-6 Astra → · Claude Fable 5.1 → · Gemini 3.8 Flash →
The story so far
3 episodes →- Google rolls out Genie 4 world model to trusted testersthis story
Import AI 479: Spacewar; low-latency tasks versus LLMs; Genie 4
Import AI · 2 November 2026
Loading the full article…
This text was published by Import AI and written by Jack Clark. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Generative AI & Models
All →- Anthropic reportedly prioritizes safety over image generation for Claude · 1 src
- qwenGPU compiles a GPU into Qwen3.5-2B weights · 2 src
- JetBrains/Mellum2.1-12B-A2.5B-Thinking-MTP · Hugging Face · 1 src
- Google releases Gemini 4 Carbon for cyber defense · 1 src
- Amazon launches Alexa+ early access in Japan with generative AI for Prime members · 4 src
Comments
via GitHub Discussions