Jev just got popular, and a domestic 29B open-source model has come to challenge it! Not only can it make judgments, it can also directly fill in parameters.

What happened: China Telecom's XingChen AGI lab open-sourced Xing4.0-29B-Intent, a 29B model that combines intent classification with parameter extraction, releasing the model, training corpus, and inference SDK. Key numbers: 29B parameters; released October 19; end-to-end latency under 150ms; zero-shot Micro-F1 of 70.27, 12.62 points above Jev's best Why it matters: The release targets the gap left after intent detection, converting natural language into structured parameters. The content frames this as a difference from Jev, which handles fast decisions within predefined options.

Domestic 29B open-source model challenges Jev with judgment and parameter-filling capabilities.

Domestic 29B open-source model challenges Jev with judgment and parameter-filling capabilities. — impact context not separately stored; see What changed.