Back to Models
XVERSEProprietary

XVERSE-Ent-A4.2B

XVERSE-Ent-A4.2B is a foundation model developed by Yuanxiang XVERSE. It has approximately 250.0B parameters and supports an 8K context window.

Parameters

250.0B

Context Window

8K

License

FREE-COMMERCIAL

Release Date

2025-12-30

API Pricing

API pricing for this model is not yet available

Strengths

  • Large 250B parameter scale
  • Free for commercial use
  • Strong performance as a foundation model

Weaknesses

  • Context window limited to 8K
  • Closed-source licensing model
  • Lacks detailed benchmark data

Use Cases

  • Commercial application development
  • General tasks with large-scale data
  • Building AI using foundation models

Deep Analysis

Architecture

MoE (25B total, 4.2B active)

Entertainment-domain Chinese model

Language

Chinese

Context Length

8K tokens

Layers

28

Experts

64 non-shared + 2 shared

8 selected per token

Specialization

Entertainment (角色扮演, 小说, 游戏)

Role-play, fiction, games

Release Date

December 2025

Strengths

  • First open-source entertainment-domain LLM
  • Strong role consistency and long-story understanding
  • 4.2B active parameters for efficient inference
  • Single-card deployment supported
  • Retains 98%+ general capabilities

Weaknesses

  • 8K context is very limited
  • Narrow entertainment domain focus
  • Chinese-only model
  • Small community (10 GitHub stars)

Competitor Comparison

ModelArenaSWEGPQAPrice
XVERSE-Ent-A5.7B---Free (open)
Qwen 2.5 7B---Free (open)
General purpose 7B models---Free (open)

Analysis generated: 2026-05-24