01 Top storiesPrimary intelligence AI & ModelsStory 01 of 03
Google launches Gemini 4 Hydrogen for agentic work Image: blog.google Google introduced Hydrogen as an efficient Gemini 4 model for coding and agent workflows, and described multi-model orchestration in Antigravity.
Why it matters Faster lower-cost worker models can change the economics of large agent systems.
AI & ModelsStory 02 of 03
Gemini Decision 1 targets fast, bounded AI decisions Image: blog.google Google introduced a Decisions API for classifying or scoring many closed-choice questions in a single multimodal pass, instead of generating lengthy text responses.
Why it matters A specialized classifier can reduce cost and latency for tool routing, policy checks, and evaluation.
AI & ModelsStory 03 of 03
Runway releases open-weight Praxis-1 robotics world-action model Image: runway.com Runway opened the weights and adaptation code for Praxis-1, a 27B-parameter world-action model trained for robotic control and scene dynamics.
Why it matters Open-weight robotics models can lower barriers for adapting vision-action systems to specialized hardware.
02 Also worth knowingRapid scan briefs Anthropic launches Claude Fable 5.2 A new demanding-work model was announced earlier in the week; the Nov 12 report revisited its coding and research claims, which require independent replication.
AI & Modelsdeepmind.google Gemini 4 Carbon reaches broader availability Google's higher-capability Gemini 4 Carbon remained an important reference model for comparisons with the newer Hydrogen.
AI & Modelsdocs.qwencloud.com Qwen4-Max preview opens Alibaba's Qwen4-Max-Preview is accessible in a limited preview period with multimodal and reasoning capabilities; access conditions matter.
← Full November 12 edition · All recent AI & Models stories