LLM EXPLORER 59,420 MODELS INDEXED

MiniCPM S 1B Sft by openbmb

By openbmb · 666 downloads

MiniCPM S 1B Sft is an open-source language model by openbmb. Features: 1b LLM, VRAM: 2.7GB, Context: 4K, License: apache-2.0, LLM Explorer Score: 0.13.

  Arxiv:2310.04564   Arxiv:2312.12456   Arxiv:2402.03804   Arxiv:2402.13516   Arxiv:2404.06395   Conversational   Custom code   En   Minicpm   Modelbest   Pytorch   Region:us   Safetensors   Thunlp   Zh

MiniCPM S 1B Sft Parameters and Internals

Model Type 
text generation, multimodal
Supported Languages 
en (high), zh (high)
Training Details 
Data Sources:
Mixed Dataset
Data Volume:
473.02 billion tokens
Methodology:
ProSparse
Training Time:
101,000 steps
Model Architecture:
Swish-activated LLaMA2 to ReLU-activated
Input Output 
Input Format:
<η”¨ζˆ·>{prompt}, Ensure bos token `~~` in inputs
Accepted Modalities:
text
Output Format:
text response
Performance Tips:
Use standard chat prompts
Release Notes 
Version:
1.0
Date:
2024-04-15
Notes:
Initial release with ProSparse approach and high sparsity.
LLM NameMiniCPM S 1B Sft
Repository πŸ€—https://huggingface.co/openbmb/MiniCPM-S-1B-sft 
Model Size1b
Required VRAM2.7 GB
Updated2026-08-05
Maintaineropenbmb
Model Files  2.7 GB   2.7 GB
Supported Languagesen zh
Model ArchitectureMiniCPMForCausalLM
Licenseapache-2.0
Context Length4096
Model Max Length4096
Transformers Version4.36.0
Tokenizer ClassLlamaTokenizer
Vocabulary Size73440
Torch Data Typebfloat16

Best Alternatives to MiniCPM S 1B Sft

Best Alternatives
Context / RAM
Downloads
Likes
BitCPM4 1B32K / 2.7 GB7526
MiniCPM 1B Sft Bf164K / 2.7 GB1536519
ProSparse MiniCPM 1B Sft4K / 2.7 GB4033
Note: green Score (e.g. "73.2") means that the model is better than openbmb/MiniCPM-S-1B-sft.