---
title: OpenAster1-4k-base
canonical_url: "https://www.modelscope.cn/models/TYFTYF/OpenAster1-4k-base"
md_url: "https://www.modelscope.cn/models/TYFTYF/OpenAster1-4k-base.md"
repository: TYFTYF/OpenAster1-4k-base
last_updated: 2026-06-25
license: mit
model_type:
  - qwen3_moe
architectures:
  - Qwen3MoeForCausalLM
parameters: 2.0B
tensor_type:
  - BF16
library_name:
  - safetensors
downloads: 26
stars: 0
tags:
  - OpenAster1
  - text-generation
  - moe
---

# OpenAster1-4k-base

> OpenAster1-4k-base - TYFTYF 在 ModelScope 开源的模型。OpenAster1 4K base model before long-context extension

TYFTYF/OpenAster1-4k-base 是 ModelScope 魔搭社区上的 2.0B 参数机器学习模型，采用 mit 许可。

- **Repository**: TYFTYF/OpenAster1-4k-base
- **License**: mit
- **Parameters**: 2.0B
- **Tags**: OpenAster1, text-generation, moe
- **Downloads**: 26
- **Stars**: 0
- **Last updated**: 2026-06-25

Source: https://www.modelscope.cn/models/TYFTYF/OpenAster1-4k-base

---

# OpenAster1-4k-base

OpenAster1-4k-base is the 4K text base model before context extension from the OpenAster1 release. This checkpoint preserves the original 4K context setting before the YaRN long-context extension stages.

The release contains model weights, tokenizer files, and generation/config files required for inference. Training-state files and optimizer states are intentionally omitted.

## Links

- Hugging Face: https://huggingface.co/binichallein/OpenAster1-4k-base
- ModelScope: https://www.modelscope.cn/models/TYFTYF/OpenAster1-4k-base
- Code: https://github.com/binichallein/OpenAster-1
- Data: https://huggingface.co/datasets/binichallein/OpenAster-1-data