The domestic large model ecosystem has added a major new member. Xunzhang Technology (XVERSE) officially open-sourced its foundation large model for the general entertainment field, XVERSE-Ent, along with Chinese and English versions. This model is deeply optimized for core scenarios in general entertainment, such as social interaction, game storytelling, and cultural creation (including novels, scripts, short video scripts, etc.), supporting lightweight deployment and rapid vertical field implementation. It becomes the first dedicated open-source large model for the general entertainment industry in China, filling the gap in high-quality foundational models in this field.

 Designed for "fun, useful, and collaborative"

Differing from general large models that pursue broad knowledge coverage, XVERSE-Ent builds around the core needs of general entertainment users, from training data, instruction fine-tuning to evaluation systems:

- Social Interaction: excels at generating natural, interesting, emotionally charged dialogues, suitable for virtual characters, AI chat companions, and community content generation;

- Game Storytelling: can automatically generate mission plots, NPC dialogues, and world settings, supporting dynamic story engines with multiple branches and endings;

- Cultural Creation: possesses strong coherence, style imitation, and rhythm control capabilities in long-text creation such as novels, scripts, and anime scripts.

The model integrates a large amount of Chinese online literature, script libraries, game dialogues, and multilingual film and television texts during training, ensuring the content aligns with local cultural contexts while possessing international expression capabilities.

 Lightweight + Open Source, Lowering the Barriers for General Entertainment AI

XVERSE-Ent emphasizes deployment friendliness and ecological openness:

- Provides multiple parameter versions such as 7B and 13B, which can run on consumer-level GPUs or edge devices;

- Uses a commercially friendly open-source license, allowing developers to use it free of charge for commercial products;

- Offers open scenario-based fine-tuning templates and evaluation toolkits, helping game companies, content platforms, and creators quickly integrate them.

Xunzhang stated that the goal of XVERSE-Ent is to become the "AI content engine" for the general entertainment industry, enabling small teams to have intelligent generation capabilities comparable to leading manufacturers.

 AIbase Observation: The Era of "Scenario-Specific" Large Models Has Begun

After vertical large models emerged in fields such as finance, healthcare, programming, and education, the general entertainment sector—a global industry exceeding $1 trillion—has finally welcomed its own AI foundation. The release of XVERSE-Ent marks that domestic large models are shifting from "large and comprehensive" to "specialized and refined," replacing parameter scale with scenario depth, becoming a new competitive focus.