LLM EXPLORER 58,531 MODELS INDEXED

DASD 4B Thinking by Alibaba-Apsara

By Alibaba-Apsara · 230 downloads
HF Repo ↗ Copy URL

DASD 4B Thinking is an open-source language model by Alibaba-Apsara. Features: 4b LLM, VRAM: 8.1GB, Context: 256K, License: apache-2.0, LLM Explorer Score: 0.25.

  Arxiv:2512.20908   Arxiv:2601.09088   Conversational Dataset:alibaba-apsara/superio...   Deploy:azure   Endpoints compatible   Qwen3   Region:us   Safetensors   Sharded   Tensorflow
Model Card on HF 🤗: https://huggingface.co/Alibaba-Apsara/DASD-4B-Thinking 

DASD 4B Thinking Parameters and Internals

LLM NameDASD 4B Thinking
Repository 🤗https://huggingface.co/Alibaba-Apsara/DASD-4B-Thinking 
Model Size4b
Required VRAM8.1 GB
Updated2026-08-15
MaintainerAlibaba-Apsara
Model Typeqwen3
Model Files  5.0 GB: 1-of-2   3.1 GB: 2-of-2
Model ArchitectureQwen3ForCausalLM
Licenseapache-2.0
Context Length262144
Model Max Length262144
Transformers Version4.51.3
Tokenizer ClassQwen2Tokenizer
Padding Token<|endoftext|>
Vocabulary Size151936
Torch Data Typebfloat16
Errorsreplace

Best Alternatives to DASD 4B Thinking

Best Alternatives
Context / RAM
Downloads
Likes
FastContext 1.0 4B SFT256K / 8.1 GB5735357
GRPO 4 70256K / 8.1 GB120
Qwen3 4B Instruct 2507256K / 8.1 GB3109972911
Fable Traces256K / 8.1 GB436210
Lightning 4B256K / 8.1 GB136
FastContext 1.0 4B RL256K / 8.1 GB455961
...PT Gpt5.1 Thinking Safetensors256K / 8.1 GB1162
Qwen3 4B Thinking 2507256K / 8.1 GB291023606
...pt Full Qwen3 4b Instruct 2507256K / 8 GB3651
Qwen3 4B Instruct 2507 FP8256K / 5.2 GB110074478
Note: green Score (e.g. "73.2") means that the model is better than Alibaba-Apsara/DASD-4B-Thinking.