Google Releases Gemini 4 Neon Model with Multi-Agent Collaboration Support
18 hour ago / Read about 0 minute
Author:小编   

Google has launched its new AI model, Gemini 4 Neon, which features a million-token context window and can process multi-format materials including spreadsheets, PDFs, charts, and audio/video, piecing together fragmented information to generate reports. In coding scenarios, the model supports cross-file troubleshooting and code refactoring, and leverages the Google Antigravity development environment to orchestrate multiple AI sub-agents for collaborative work: it handles task scheduling, while the Hydrogen sub-agent drives work forward in parallel and tracks progress on long-running tasks—making it well suited for multi-step, continuous workflows such as large codebase troubleshooting and massive data analysis. In terms of performance, the model ranks third on the Artificial Analysis Intelligence Index at the xhigh (highest) reasoning tier and tops the Programming AI Agent Index with a score of 58.6. At the default high tier, the cost per task is 32% lower than that of its sibling, Gemini 4 Carbon. Early internal evaluations show that its overall task quality is on par with the larger Gemini 4 Carbon, while its processing speed can reach three times that of the latter.