LLM EXPLORER 60,277 MODELS INDEXED

Crystal Think V1.0 by sizzlebop

By sizzlebop · 52 downloads
HF Repo ↗ Copy URL

Crystal Think V1.0 is an open-source language model by sizzlebop. Features: 4b LLM, VRAM: 8.1GB, License: apache-2.0, LLM Explorer Score: 0.18.

  Autotrain compatible Base model:adapter:qwen/qwen3-...   Base model:qwen/qwen3-4b   Conversational Dataset:nvidia/openmathreasoni...   En   Endpoints compatible   Finetuned   Grpo   Lora   Math   Mathematical-reasoning   Qwen3   Reasoning   Region:us   Safetensors   Sharded   Tensorflow
Model Card on HF 🤗: https://huggingface.co/sizzlebop/crystal-think-v1.0 

Crystal Think V1.0 Parameters and Internals

LLM NameCrystal Think V1.0
Repository 🤗https://huggingface.co/sizzlebop/crystal-think-v1.0 
Base Model(s)  Qwen/Qwen3-4B   Qwen/Qwen3-4B
Model Size4b
Required VRAM8.1 GB
Updated2025-06-24
Maintainersizzlebop
Model Files  0.3 GB   5.0 GB: 1-of-2   3.1 GB: 2-of-2
Supported Languagesen
Model ArchitectureAutoModelForCausalLM
Licenseapache-2.0
Model Max Length40960
Is Biasednone
Tokenizer ClassQwen2Tokenizer
Padding Token<|vision_pad|>
PEFT TypeLORA
LoRA ModelYes
PEFT Target Modulesup_proj|k_proj|v_proj|gate_proj|q_proj|o_proj|down_proj
LoRA Alpha64
LoRA Dropout0
R Param32
Errorsreplace

Best Alternatives to Crystal Think V1.0

Best Alternatives
Context / RAM
Downloads
Likes
Qwen 3 4B Text To SQL0K / 8.1 GB146
Crystal Think V20K / 8.1 GB177
Crystal Think V10K / 8.1 GB521
Phi 3 Bangla Instruct0K / 0.5 GB85
MING MoE 4B0K / 0.3 GB141
Note: green Score (e.g. "73.2") means that the model is better than sizzlebop/crystal-think-v1.0.